Key takeaway
When a downstream integration goes quiet, check the source before you debug everything downstream. Confirm Salesforce actually published the event you expected, with the payload you expected — one check at the producer saves a day of chasing symptoms through middleware, queues, and connectors that were working fine all along.
A downstream system goes quiet. The instinct is to open the middleware, the queue, the connector — anything but Salesforce. Nine times out of ten, that's the wrong place to start.
Before you trace the rest of the integration, confirm the source of truth: did Salesforce actually publish the event you expected, with the payload you expected?
Tip
Verify the producer before you debug every consumer. One check at the source saves hours of chasing symptoms downstream.
Watch what the source actually sent
Have someone subscribe to the event channel directly and re-trigger the action that should fire it. If nothing comes across, the problem is upstream — a flow that didn't run, a validation rule that blocked the record, a limit no one saw. If it does come across, compare it to what you assumed: a status field that's blank, an amount stored as text, a correlation id that never got set. Field names, data types, and nulls are where assumptions quietly die. (For the exact mechanics — replay clients, the trigger chain, the redelivery gotchas — see the Salesforce–ERP architecture teardown.)
Example platform event payloadJSON · example-platform-event.jsonDownloadThen, and only then, look downstream
Once you've confirmed the event published with a clean payload, you've cut the problem in half. Now the question is narrow: why didn't the consumer act on a message you can see with your own eyes? Replay id gaps, subscriber errors, and retry storms all live here — but you're debugging one system instead of five.
Watch out
Platform events aren't guaranteed to be delivered forever. If your subscriber was down past the retention window, the event is gone — reconcile from the records, don't wait for a redelivery that isn't coming.
This is the whole philosophy behind an Integration Spine™: every system exposes what it did, so you can trust the flow instead of guessing at it.
Related tipFormAssembly → Salesforce: the mapping mistake that silently drops leadsA practical next step
If this looked familiar, check your own Integration Spine.
See where systems, data, ownership and controls may have stopped connecting.
More Bytes
Integration & RevOps7 hard lessons from syncing Salesforce to an ERP
Each of these was paid for in production — including the afternoon a redelivery storm quietly doubled receipts in the general ledger while every dashboard stayed green.
Integration & RevOpsFormAssembly → Salesforce: the mapping mistake that silently drops leads
Your form submits, the connector says success, and the lead never lands. Here's the field-mapping trap that swallows records without an error.
PlatformsYour CRM isn't the problem — your lifecycle definitions are
Before you migrate platforms or blame the tool, get four teams to agree on what a lead, an MQL, and an opportunity actually mean.