Troubleshooting — by what you see

Integration Core help · for the operator and support · graded against Core v1.0.0-rc.5 · written 2026-10-06 · re-read against rc.8's screens 2026-10-09 · where a newer Core words a screen differently, see the Glossary's "Changed in newer releases"

This page covers the symptoms every integration shares. Each entry says what you see · what it means · what to do · when to call support. For a cause particular to one system — its keys, its own errors, its own screens — read that connector's own troubleshooting page as well.

What this page restates, and from where: the support runbook §2, §3, §5, §6, §8, §10 and §11 (2026-09-13, re-looked-up 2026-10-02); Integration Core v1.0.0-rc.6 — known issues §1, §2 and §5 (2026-10-02); the setup guide, Steps 5–7 and Getting alerts by email (screens re-derived 2026-09-27). Each entry ends with its source line. Where this page and a paper disagree, report it; a section that read a newer build than its paper says so.

Find your symptom

You see	Go to
A connection reads Not reported yet	Not reported yet
A connection reads Not checked for …	Not checked for a while
… refused the sign-in	The other system refused the sign-in
Cannot reach …	The other system cannot be reached
Stopped part-way, or an empty Stopped part-way list	Stopped part-way
A connection is Switched off and nobody remembers doing it	A connection switched itself off
"This connection cannot be switched on"	Switch on is refused
Nothing has crossed since switching on	Nothing has happened since switching on
A record reads Unverified	A record says Unverified
A record reads Failed	A record failed
A record reads Cannot be sent	A record cannot be sent
There is no Retry button	There is no Retry button
Retry selected skipped records, or said not sent again	Retry selected skipped some records
A Needs a decision record comes back	A decision keeps coming back
Two decisions are open on one order	Two decisions are open on one order
Many records read Waiting	Many records are waiting
A backlog moves slowly after switching on	A backlog drains slowly
A money record for … is waiting for its next try	A money record is waiting for its next try
A record is Held behind an unknown attempt	Held behind an unknown attempt
A Reversal not confirmed is hours old	A reversal is not confirmed
A record you recorded by hand was sent again	A held document was also keyed by hand
A list shows fewer rows than its tile, or 0 of 3	A list shows the wrong rows or count
A money record shows no amount	A money record shows no amount
Amount Confirmed is ticked but their bill is higher	Amount Confirmed but the bill is higher
An amount or currency alert	An amount or currency alert
Too many alerts, or none	Alerts: too many, or none at all
The alert email does not arrive, or arrives twice	The alert email does not arrive
Deleting an iVendNext record is refused	Deleting a linked record is refused
A record reads Decided or Ignored	A record says Decided or Ignored
A refund the other system issued never appears	A record the other system sent never appears
A change made in iVendNext never reached the other system	A change made here has not reached the other system
A value changed and changed back, and the change back never came through	A change back to an earlier value was dropped
One order shows many Retrying rows	One order shows many Retrying rows
An alert this page does not name	An alert this page does not name
A save is refused, Ask a … fails, or a Hub tile is gone	Small things that go wrong on the screens
A document cancelled here is still live there	A cancelled document is still live there

Not reported yet

What you see. A connection reads Not reported yet, blue or grey.

What it means. It is switched on and has never once reported back. Blue: its Check Every (minutes) value plus an hour has not passed yet — nothing is wrong yet. Grey: that window has passed — it is overdue. ⛔ Never read either as running.

What to do. Blue: wait one interval. Grey: ask whoever administers your site to check that the site's scheduled jobs and the connector app's scheduled jobs are running. ⚠ Clicking the chip lists every switched-on connection that has never reported, including one switched on a minute ago — read each row's own word.

When to call support. ⛔ Nothing will alert you about this one — you have to look. If it is still grey after the scheduler is confirmed running, raise a ticket.

Source: the setup guide, Step 6 (2026-09-27); the support runbook §5 and §6 (2026-09-13).

Not checked for a while

What you see. A connection reads Not checked for N min / hours / days (red).

What it means. It was reporting, and has stopped looking for changes within its own schedule plus an hour. Anything that only arrives when it looks is not being picked up.

What to do. The same question as above: has the connector's scheduled job stopped? Ask whoever administers your site.

When to call support. If the scheduler is running and the connection still does not report.

Source: the setup guide, Step 6 (2026-09-27); the support runbook §5.

The other system refused the sign-in

What you see. System refused the sign-in since time (red) on the Connections list; the connection's own form says the connection refused the sign-in details (since …); the Hub's Your connections card lists it under Sign-in refused. Records may be Set aside.

What it means. The other system refused the credentials at the last check. Records are set aside, not failed, and nothing is lost.

What to do. Ask a System Manager to renew the credentials on the connection's setup page — the form offers Ask a System Manager to anyone else who can edit it. Work resumes by itself on the first accepted call.

When to call support. If renewed credentials are refused again, or the connector's own page says something different about its sign-in.

Source: the setup guide, Step 6 (2026-09-27); the support runbook §5 (2026-09-13).

The other system cannot be reached

What you see. Cannot reach system since time (red). After twice the connection's check interval (never less than 150 minutes), the bell rings "Cannot reach …".

What it means. The other system did not answer. Work is paused, not lost, and not retried against it.

What to do. Ask the other system's owner whether it is up, and whether its address changed. ⚠ A short outage can still ring the bell on some connectors; read the connection's Cannot Reach It Since before escalating.

When to call support. If the other system is up and answering, and the connection still reads Cannot reach.

Source: the setup guide, Step 6 (2026-09-27); the support runbook §5 and §6; the known issues §5 (2026-10-02).

Stopped part-way

What you see. A connection reads Stopped part-way (red), or the Stopped part-way link is empty.

What it means. Its runs report in, but the connector recorded that its last run stopped before finishing — a far side that moved, a firewall, an answer it could not read. Integration Core is told it stopped, never why.

⚠ Only a connector that reports this can ever show it. Acumatica reports it from its tag v1.0.0-beta.3 (3242c82, controllers/drive.py::_note_pass_completed); WooCommerce does not, at its v1.0.0-beta.3; an older Acumatica build does not either. So an empty list means none reported, ⛔ not proven fine. Until your connector reports it, a connection with no new activity during trading hours is suspect.

What to do. Check the other system first. A System Manager may find the reason in the Error Log, if the connector wrote one.

When to call support. Then ask whoever supports the connector.

Source: the setup guide, Step 1 and Step 6 (2026-09-27); the support runbook §5; the known issues §2, BUG-IC-003 (2026-10-02); Acumatica's troubleshooting page (beta edition, 2026-10-05).

A connection switched itself off

What you see. A connection reads Switched off, and the line naming who last changed the switch says Administrator — often right after an upgrade.

What it means. The system switched it off and signed its own decision. After an upgrade, a difference in what the connector itself declares (its class, its contract version, its identity) switches the connection off and alerts — the app is not the app that registered.

What to do. ⛔ Do not simply switch it back on. Read the alert, which names what changed, and ask a System Manager to read the Error Log. ⛔⛔ If you are told to re-run the connection's setup, run it against the same system — never point an existing connection at a different store or tenant (that keeps every record link, so money documents can be matched to the wrong system's IDs). For a different system, create a new connection.

When to call support. Always, with the alert's text.

Source: the support runbook §6 (switched_off) and §7 (2026-09-13); the known issues §2, IC-DP7-B5 (2026-10-02).

Switch on is refused

What you see. "This connection cannot be switched on".

What it means. Usually another connection of the same kind is already running, and only one is allowed (an ERP: one per company). The message names what is in the way.

What to do. Switch that one off first — a switched-off connection does not hold the place.

When to call support. If nothing of the same kind is running.

Source: the setup guide, Step 5 (2026-09-27); the support runbook §10.

Nothing has happened since switching on

What you see. The connection is on, and no record has appeared in Integration Activity.

What it means. Nothing runs until the connector's own scheduled check runs for the first time. Until then the connection reads Not reported yet.

What to do. Check the site's scheduler and the connector's scheduled jobs before treating it as a data problem. The only evidence an integration is working is a real record reaching Applied.

When to call support. If the scheduler runs and the connection still has not reported after its interval plus an hour.

Source: the support runbook §8 item 3 (2026-09-13); the setup guide, Step 5.

A record says Unverified

What you see. A record reads Unverified (the Hub tile Cannot tell if it arrived); the record says "Integration Core cannot tell whether this reached …".

What it means. ⭐⭐ We cannot tell whether it reached the other system. It may have landed. It is not failed and not pending.

What to do. ⛔⛔ Nothing. Never retry it, never try to force it. Only the other system can settle it — the connection's next status check, or a person there; the record's Error says which, where the connector wrote one. A burst right after go-live usually settles on the next status check.

⚠ On Acumatica, a record may ask you to release a document in Acumatica. With an Acumatica build older than its tag v1.0.0-beta.3 (which carries the fix, 3242c82), that document's total was not compared first — and a record such a build already named keeps its sentence after an upgrade. Before releasing any document an Acumatica record asks you to release, look for an amount or currency alert on it ("Amount differs: …", "Currency mismatch: …" or "Currency not comparable: …"); if there is one, do not release it as it stands — correct or void it in Acumatica first. Acumatica's own troubleshooting page, A document is waiting to be released, has the steps.

When to call support. If it is still there the next day. Escalate; the answer is on the other system, not here.

Source: the support runbook §2 and §8 (2026-09-13); the setup guide, Step 7 item 2 (2026-09-27); the known issues §2, IC-ECHO-RULE-B1 and IC-RELEASE-RC7-B3 (2026-10-06).

A record failed

What you see. A record reads Failed; it says "This record failed. Error: …".

What it means. It was tried and rejected, and parked for a person.

What to do. Read the error. Press Open … to reach the document, fix the cause, then press Retry. ⛔ Do not retry without reading the error. ⚠ The Error column on the list is only drawn on a wide window — open the record to read it.

When to call support. If the error does not say what to fix, or the same error returns after the fix.

Source: the support runbook §1 and §2 (2026-09-13); the setup guide, Step 7 item 1 (2026-09-27).

A record cannot be sent

What you see. A record reads Cannot be sent (red) — "The document this change belonged to no longer exists…".

What it means. The iVendNext document is gone, so the other system can never be told. No retry can help.

What to do. Fix it on the other system by hand if it matters, then press Acknowledge. It sends and recovers nothing; it records that you saw it and stops the hourly alert. The record then reads Acknowledged.

⚠ A row that reads Failed and offers Acknowledge but no Retry is the same in effect: the connection marked it as a failure no retry can fix. Fix it on the other system if it matters, then acknowledge it.

When to call support. ⚠ If a Cannot be sent row shows no Acknowledge button, your role may not edit it — ask an Integration Manager. If a row only reads like a permanent loss (a Failed row whose error says the document is gone) it offers Retry but no Acknowledge: the connector failed it without marking it. ⛔ Do not retry it; raise a ticket for the connector team.

Source: the support runbook §3 (2026-09-13); the setup guide, Step 7 item 1 (2026-09-27); the Acknowledge button read at Core rc.5.

There is no Retry button

What you see. A record has no Retry button.

What it means. One of these: its state refuses a retry (nine states do, each for its own reason); it is a failure no retry can fix or a lost document (Cannot be sent); it is a test record; its connection is switched off; or your role may not edit the record. The record's notice at the top says which reason applies.

What to do. Read the notice first. ⛔ There is no override for the nine, for anyone. ⚠ The opposite also holds: a Waiting, Set aside, Decided or Ignored record does show Retry. ⛔ Do not press it on Held behind an unknown attempt; after retrying a Decided record, follow it on Integration Activity until it finishes.

When to call support. If the notice does not explain it.

Source: the support runbook §2, Which states refuse a retry (2026-09-13); the Retry button read at Core rc.5; the known issues §2, IC-MONEY-SHOWN-B1 (2026-10-02).

Retry selected skipped some records

What you see. Retry selected lists records it will skip, or afterwards says "… were not sent again" in orange, or lists records as not attempted.

What it means. Skipped records are in a protected state (each with its reason). Not sent again means the connector declined them — correct behaviour, not a fault. Not attempted means they were never tried.

What to do. Read the not attempted list: those still need a retry. Confirm the outcome from each record's state, not from the message.

When to call support. If a record is declined every time and you cannot see why.

Source: the support runbook §2 and §10 (2026-09-13); the setup guide, The things you can press.

A decision keeps coming back

What you see. A Needs a decision record you closed with Keep iVendNext Version appears again.

What it means. Closing it records the decision; it does not remove the cause. On a connector that polls, the next pass can raise a fresh one. A customer saying "it keeps coming back" is describing known behaviour.

What to do. Remove the cause on the other system. ⚠ On WooCommerce: an order cancelled and then restored on the store leaves an open decision about the cancellation that no longer applies — check the order's current status on the store, then close the decision by hand.

When to call support. If the cause is removed and it still returns; give the order number.

Source: the support runbook §2 and §10 (2026-09-13, re-looked-up 2026-10-02); What Integration Core gives you today, Live defects.

Two decisions are open on one order

What you see. A decision record says "Also open on the same record: …".

What it means. The other system changed the same order twice before anyone answered. Each question is named on the other's record. Integration Core does not say which is current; two questions can be different money questions.

What to do. Read every one first. Where they are the same question, answer the newest and leave the others.

When to call support. If the questions disagree about money and you cannot tell which is right.

Source: the support runbook §2 (2026-09-13); the known issues §2, BUG-IC-056 (2026-10-02).

Many records are waiting

What you see. A long Waiting on a setup change list.

What it means. Waiting is not failing. One missing piece of setup often holds every record of that kind.

What to do. Open one row and read its Waiting on: reason. Use Open the fix where it is offered, fix the one thing, and they all move. Two exceptions: a backlog after a switch-off (it is on Kept while switched off and drains by itself), or a missing record link (use Record Links → Link manually, then retry the record).

When to call support. If the same record still waits a day after you fixed what it named.

Source: the support runbook §8 item 1 (2026-09-13); the setup guide, Step 7 (2026-09-27).

A backlog drains slowly

What you see. After a connection is switched back on, records drain for hours.

What it means. After the first batch, the re-queue moves 500 records an hour. Twenty thousand take about forty hours. That is by design; the records are safe while they wait.

What to do. Wait. ⛔ Do not diagnose this as a stuck queue.

When to call support. If the count stops falling for more than an hour while the connection is on.

Source: the setup guide, Step 5 (2026-09-27); the support runbook §10.

A money record is waiting for its next try

What you see. A bell: A money record for ⟨connection⟩ is waiting for its next try. The record's Attempt No has not moved.

What it means. The connection's retry pass has not tried this record again for three turns of its schedule (four hours at least; three days on a once-a-day connection). It may be waiting behind other records, on a run that stopped (a refused sign-in, a busy other system), or the retry pass may not be reaching it. The connection retries it, not Integration Core. The bell names the oldest such record and counts the rest.

What to do. Open the record from the bell. If the connection's chip shows a refused sign-in or Cannot reach, fix that first; the record goes on its next try. A long backlog after an outage clears itself. Otherwise send the record's reference and Attempt No to the connector team.

When to call support. If the bell rings again for the same record a day later with nothing on the connection's chip.

Source: the support runbook §10 (2026-10-09); the known issues §5, IC-REDRIVE-ONE-ROW-B4 (fixed in rc.8).

Held behind an unknown attempt

What you see. A record on Held behind an unknown attempt.

What it means. ⭐⭐ An earlier attempt at the same record may already have reached the other system. Sending this one could duplicate it.

What to do. ⛔ Do not force it, and do not press the Retry button its record shows. Wait for the status check to settle the earlier attempt.

When to call support. If it stays held for a day.

Source: the support runbook §2 and §6 (held_behind_unknown) (2026-09-13); the Retry button read at Core rc.5.

A reversal is not confirmed

What you see. A record on Reversals not confirmed (it reads Reversal not confirmed); after four hours the bell rings.

What it means. ⭐⭐ iVendNext reversed its side and the other system has not. On a refund, the customer's money has not left yet.

What to do. Treat it as money owed. Retry is refused — sending again could reverse twice.

When to call support. If one is more than a few hours old.

Source: the setup guide, Step 7 item 4 (2026-09-27); the support runbook §2 and §6.

A held document was also keyed by hand

What you see. A document a person recorded by hand on the other system appears there twice.

What it means. A connector's held for a person record waits as Waiting. When it is re-tested and its cause has cleared, it is sent — so a copy keyed by hand becomes a second document.

What to do. ⛔ Never key by hand into the other system a document the Hub still shows as Failed, Waiting or Unverified without first reading its record. If you must record it by hand, follow the connector's own page for how to stop the original being sent.

When to call support. At once, with both documents' numbers.

Source: the Acumatica team's ⚑ of 2026-10-05 (AC-DOCS-1, their AC-B215) and Acumatica's troubleshooting page (beta edition, 2026-10-05).

A list shows the wrong rows or count

What you see. A list opened from a Hub card shows fewer rows than you expect, or the count at the top right reads 0 of 3 with three rows showing.

What it means. ⚠ On rc.5, Connections and Integration Activity opened from a Hub card reopen with the filters you last used there — after a tile click, still that tile's filters. Separately, the N of M count can be wrong for a moment after a filter changes.

What to do. Read the filter pills under the title and clear them (×). Press the list's refresh button before trusting the number or the rows.

When to call support. If the rows are still missing with every filter cleared.

Source: the known issues §5, IC-HUB-CARDS-B2 (2026-10-02); the setup guide, Step 6 (2026-09-27).

A money record shows no amount

What you see. A money record reads "No amount on this record".

What it means. ⛔ Not zero: the connector did not record an amount of ours.

What to do. Read the other system's own figure — Amount as Received where the connector passed it, or the record on the other system.

When to call support. If the other system's figure and ours cannot be reconciled.

Source: the setup guide, Step 7 item 5 (2026-09-27); the support runbook §10; the known issues §3 (2026-10-02).

Amount Confirmed but the bill is higher

What you see. A record has Amount Confirmed ticked, and the other system's invoice is higher.

What it means. Amount Confirmed means the amount the other system reported back for what crossed matched iVendNext's. ⛔ It is not their final bill: a system that adds its own tax or fee can bill more while the tick stands. A record can show Final Bill in Connected System only where the connector reports it, and at rc.5 neither shipped connector does.

What to do. Check the bill on the other system itself. A blank Final Bill means not reported, ⛔ never no difference.

When to call support. If the difference is not a tax or fee the other system adds.

Source: the setup guide, Step 7 (2026-09-27); the support runbook §2 (2026-09-13).

An amount or currency alert

What you see. The bell reads "Amount differs: … reported, … here", "Currency mismatch: …" or "Currency not comparable: …".

What it means. The amount check did not pass when the record was closed. A record that was waiting or needed a decision keeps its state and its amount.

What to do. Read the record's amount lines (Reported by …: …, not …). ⚠ On an outbound waiting record, settle the money question before fixing the wait — once the wait clears, the connector may send it. On an Acumatica record that asks you to release a document, do not release it (see A record says Unverified). To find these alerts, search the bell for :amount: and :currency not comparable:.

When to call support. Escalate to the connector team with both figures.

Source: the support runbook §6 (amount, currency not comparable) (2026-09-13, re-looked-up 2026-10-02).

Alerts: too many, or none at all

What you see. Hourly alerts about a connection you retired; or a queue nobody acted on and no alerts.

What it means. Alerts are one per connection, per kind, per hour — the number of alerts says nothing about the number of records. A connection you switched off does not ring for the records waiting for it, ⚠ but its money records still ring (unverified, unconfirmed reversals, decisions, failures). If nobody holds an enabled Integration Manager or System Manager role, nothing is written at all; if the scheduled jobs stop, nothing rings.

What to do. Too many: answer or acknowledge each money record (its form names the way), or switch the connection back on. None: check who holds the Integration Manager role, and that the scheduler runs — and, from rc.8, whether a User Permission on the Connection limits that Manager to other connections.

When to call support. If both are right and alerts still do not arrive.

Source: the support runbook §6 (2026-09-13); the setup guide, Step 7 (2026-09-27).

The alert email does not arrive

What you see. Alerts reach the bell but no email arrives; or each arrives twice; or a subject shows &.

What it means. Email needs an outgoing email account first (a System Manager's job); without one, every alert writes an Error Log entry. Two switched-on copies email each alert twice. Alerts raised during an upgrade's data patches are never emailed, and Administrator gets no alert email. ⚠ On rc.5 the shipped Notification prints & as & and, while outgoing mail fails, re-emails each failed send — rc.6 corrects both.

What to do. Set up the outgoing email account, and keep exactly one switched-on copy of the shipped Notification. ⛔ Up to rc.7, on a site where each storefront has its own operator, leave the email off: every alert reaches every Integration Manager. From rc.8 a connection's alert and its email go to the Integration Managers who may see that connection (if none may, to the System Managers); a site-wide notice still reaches every Integration Manager.

When to call support. If the copy is refused when you try to switch it on — that has been measured only in developer mode.

Source: the setup guide, Getting alerts by email (2026-09-27); the known issues §5 (2026-10-02).

Deleting a linked record is refused

What you see. Deleting an iVendNext record is refused, naming a connection and the other system's ID.

What it means. Another system is linked to it, and it will send changes for it. ⛔ Making the link inactive does not change the refusal.

What to do. Disable the record instead of deleting it. Merging two linked records is refused too; make one link inactive first, then merge.

When to call support. If the record must really go.

Source: the support runbook §10 (2026-09-13).

A record says Decided or Ignored

What you see. A record reads Decided or Ignored.

What it means. ⭐ Closed on purpose, not failed. Decided: a person closed it with Keep iVendNext Version (their name, the time and their note are on it). Ignored: the connection closed it — for example the record type is switched off, or a delete arrived for something never linked.

What to do. Nothing. If the cause changed, a decided record may offer Re-check. ⚠ A decided record also shows Retry; if you press it, follow the record on Integration Activity yourself until it finishes — neither the bell nor Money records waiting will show it.

When to call support. Not needed.

Source: the support runbook §2 and §8 (2026-09-13); the known issues §2, IC-MONEY-SHOWN-B1 (2026-10-02); the Retry button read at Core rc.5.

A record the other system sent never appears

What you see. A refund or order the other system shows as issued never appears in Integration Activity, or its close reads "already imported as …".

What it means. ⛔ Money, and silent. The other system handed back an identifier it had used before — after a restore from backup, a sandbox clone, a re-install onto the same database or a tenant migration. Integration Core remembers the other system's IDs for ever, so the second record looks like the first arriving again: no activity row, no alert, nothing to retry.

What to do. Ask whether that system was recently restored, cloned or migrated. Compare its record dates with the Record Links creation dates for the same IDs — matching IDs with dates days apart is the signature. The money must be entered by hand once identified.

When to call support. At once, with the IDs and both dates.

Source: the support runbook §10, IC-FARSIDE-REUSE-B4 (2026-10-02); the known issues §2.

A cancelled document is still live there

What you see. A document cancelled in iVendNext is still live on the other system; a Cancelled in iVendNext difference is open.

What it means. Cancelling a document a connection already sent records a difference and rings the bell once. Nothing is sent to the other system; no connector mirrors the cancel by itself yet. ⚠ On rc.5 a link written after the cancel records no difference.

What to do. Reverse it on the other system by hand, then Mark resolved. After cancelling a document a connection sends, look it up on Record Links once its record settles: a link means it is on the other system.

When to call support. If money moved and you cannot reverse it there.

Source: the known issues §2, IC-RELEASE-RC3-B1, and Upgrading from rc.5 item 4 (2026-10-02).

A change made here has not reached the other system

What you see. A document saved in iVendNext — often in the back office — has no record on Integration Activity, or reaches the other system late.

What it means. The engine's change notification is skipped in some cases — an engine branch without the IC-X9 fix skips some back-office saves, and some saves are skipped on every branch. A connector that runs a catch-up sweep picks those up later — late, not lost. ⚠ A sweep recovers updates only: a delete, a rename, a commit that landed behind its cursor, or a sweep that has stopped are not recovered by it.

What to do. Wait one of the connection's check intervals for the sweep. A delete shows as a Missing Ours difference and a rename as Renamed.

When to call support. If it has not crossed after the next sweep, with the document name — and ask which engine branch the site runs.

Source: What Integration Core gives you today, Live defects (2026-09-13); the support runbook §10 (2026-10-02); the known issues §3, IC-R18, and §4 (2026-10-02).

A change back to an earlier value was dropped

What you see. A price, a status or a customer's details were changed on the other system and then changed back, and iVendNext stayed on the middle value. Nothing failed and nothing alerted.

What it means. ⛔ Money where the value is one. When the change back is byte-for-byte the same as the earlier one, Integration Core reads it as the earlier message arriving again, and drops it as a repeat.

What to do. Ask for the record to be edited on the other system to a state it has never held (a new value, then the value they want). ⛔ Re-saving the same value does not work, and neither does pressing Retry on an old record. ⛔ Do not edit the iVendNext record by hand.

When to call support. If the third state does not come through either.

Source: the known issues §1, IC-R22 (2026-10-02); the support runbook §10c (2026-10-02).

One order shows many Retrying rows

What you see. One web order the connection cannot import has several In progress rows — up to about 23 — all retrying, and the Error Log grows with them.

What it means. On rc.5 each time the store hands the order back, another row opens for it. They are copies of one order, not extra sales — the connection's record link stops a second invoice — and they all go Failed together at the try limit. A shopper name containing < is one known cause.

What to do. Fix or remove the cause of the order's failure; the rows then stop. ⛔ Do not press Retry on the copies.

When to call support. With the order number.

Source: the known issues §5, IC-E2E6-B1 (2026-10-02).

An alert this page does not name

The alert is about	What it means	What to do
different content arriving for a record already in flight or unconfirmed (body_changed, body_refused)	the earlier content is kept; the newer content was not applied	escalate — a genuine content dispute
a record the other system took 4 hours ago, still not final (outcome_not_final)	it is Accepted and has not completed	money: escalate if the next status check does not settle it
the connector's code or the account it runs as could not be found after an upgrade (adapter_missing, application_user_missing)	the connection cannot run	escalate
"Integration Core updated to … — open the Hub to check" (core_updated)	only that the version changed	look at the Hub
"Upgrade: Connections with a problem may be stale — see Error Log"	an upgrade could not set a flag; nothing is lost	wait for the next hourly check, then compare the tile with the Connections list
"Upgrade: daily summaries may count every connection — see Error Log" (from rc.8)	an upgrade could not stamp each daily summary with its person, so a summary made before rc.8 may still count every connection	the hourly check stamps them again; until then a person fenced to some connections should stop their daily summary
"Upgrade: a Core report email may list every connection — see Error Log" (from build 2.0.0.157)	an upgrade could not switch off the scheduled emails of Integration Core's own reports made before it, so one may still list every connection's records	a System Manager reads the Error Log and runs the migrate again (no hourly check retries this); meanwhile switch those emails off by hand
"Upgrade: the in-product help was NOT refreshed — see Error Log" (from rc.8)	the Help answers — and the help buttons that open them — may be missing or out of date (on rc.8, the six Help pages); nothing else is affected	a System Manager reads the Error Log and runs the migrate again
a step held during a migrate (role_missing, constraint_missing, workspace_stale, index_missing, …)	that step did not complete	a System Manager reads the Error Log, fixes the cause and runs the migrate again
Integration Core's own database layer changed a value the connector declared (stored_value_rewritten)	the cause is Integration Core's	nothing on the connector's side; report it

Source: the support runbook §6 (2026-09-13, re-looked-up 2026-10-02; the two rc.8 upgrade alerts read at Core d9a058f, 2026-10-09).

Small things that go wrong on the screens

• Saving a connection is refused with a bare permission message. The hourly check changed a field while the form was open. Reload the page and save again — nothing was lost.
• Ask a … is refused. Assign or message the person by hand.
• A Hub tile or link has gone. Any desk user can edit the Hub page's layout. Ask whoever administers the site to run a migrate, which draws it again; no record or count is affected.
• An alert email carries a link or an image. The only link Integration Core puts in it is Open it in iVendNext. ⛔ Follow no other link or image; open iVendNext yourself.

Source: the known issues §5, IC-HUB-CARDS-B6, IC-FIX-ROUTE-B1, IC-DP10-B20 and IC-RELEASE-RC4-B2 (2026-10-02).

What to send with a ticket

1. The connection name and its state from the Connections list.
2. The record's name, its state, its attempt count and the full error text from the form.
3. Whether the scheduler is running, and when the connection last reported.
4. The connection's Check Every (minutes) value.
5. Whether anybody holds the Integration Manager role, and whether they are enabled.
6. The relevant Error Log entries (a System Manager can read them) and bell alerts, with times.
7. The version each app is at — the Integration Core tag and the commit of each other app.
8. ⛔ Whether anything was edited by hand. Say so first; it changes the diagnosis.

Source: the support runbook §11 (2026-09-13); the setup guide, Where to raise a ticket.