I have a cloud flow using HTTP action that has been running reliably for an extended period, calling a POST endpoint and parsing the json response. In the last several runs it has started failing consistently (100% reproducible, including after 4 retry attempts with exponential backoff) with BadRequest Http request failed as there is an error: 'Error while copying content to a stream.'
No values are returned in the action's outputs, the failure occurs before a complete response can be captured, suggesting the connection is being terminated mid-transfer rather than the server returning an HTTP error status.
Request details:
URI: https://webgate.ec.europa.eu/rasff-window/backend/public/notification/search/consolidated/
Method: post
Headers: content-type: application/json
Body: small json payload (~400 bytes), unchanged from when the flow was working.
Troubleshooting already performed:
-
Disabled Allow chunking in settings - no change.
-
Added accept-encoding: identity header to rule out gzip-related stream handling - no change.
-
Added retry policy (exponential, 4 attempts) - all 4 attempts fail identically.
-
Verified server-side behavior independently: sent the identical request (same URI, headers, body) from an external tool (no cookies, no browser headers, just content-type: application/json) - received a clean 200 OK, valid json body, 29.49 KB, in under 1 second. This confirms the server is healthy and not rejecting "bare" non-browser clients.
- Confirmed this is an isolated flow with no other users or recent edits by anyone else in the tenant.
Given that an external client receives a complete, valid response to the exact same request, but the PA HTTP action consistently fails to fully read/copy that response - is this a known issue with the HTTP action's outbound network path or stream handling (e.g. specific to traffic originating from Azure datacenter IP ranges, or a recent regression in the connector)? Is there a recommended workaround other than switching to a custom connector or external proxy?