Do not resend a REST call that used up its whole deadline - #1502

Summary

A REST call that ran into its whole deadline without an answer was sent again up to three times when it was idempotent (GET, PUT). With the default 30 s timeout that is about 124 s before the model gets an error, longer than most MCP clients wait. Cloud logs of 3-4 Nov: 13 calls in 3 workspaces ended with "Could not reach the API after 4 attempts ... (ECONNABORTED)" after ~124 s (PrestaShop 11, Telegram 1, a hand-built connector 1).

Tools that raised their own timeoutMs already got no retry after a timeout. This applies the same rule to every call that used up its deadline.

Changes

• requestWithRetry: no retry when the error is the request deadline (ECONNABORTED with "timeout of ms exceeded", raised by axios and by outbound-http).
• Other ECONNABORTED errors (a socket aborted early), ETIMEDOUT, ECONNRESET, 502/503/504 and 429 are retried as before.
• The error the model sees is unchanged: "Could not reach the API: the request was aborted before the API responded (ECONNABORTED)".

Cloud and self-hosted behave the same.

Type

• Bug fix
• New feature
• Documentation
• Refactoring
• Performance improvement
• Other (describe below)

Testing

• Added/updated unit tests
• Tested manually (describe steps below)
• No tests needed (explain why)

New case: a GET and a PUT that hit "timeout of 30000ms exceeded" are sent once. The existing case "still retries a GET/PUT that timed out" (other ECONNABORTED) still passes.

In packages/backend:

• npx jest src/connectors/engines/rest.engine --maxWorkers=4 → 204 passed
• npx eslint on the changed files → exit 0
• npx tsc --noEmit -p tsconfig.json → exit 0

Checklist

• My code follows the existing code style
• I have tested my changes locally
• I have updated documentation (if applicable)
• All existing tests pass (npm test)
• This PR has a descriptive title

Related Issues

Found in the 2026-11-04 Cloud log review.