Issue Summary

Forge functions that make a request that consumes a payload may timeout and throw an error by the retry proxy if the original request threw an error.

This is fairly difficult to reproduce consistently. It requires multiple requests that may timeout.

The symptoms include

1. The Forge invoke() method starts to time out
2. Seeing HTTP 503 errors in the logs

Steps to Reproduce

1. Run a forge function in the nodejs18 runtime that makes an external request that might timeout
2. Rinse and repeat until failures start

Expected Results

The consuming stream is reset and the request is retried correctly.

Actual Results

The below exception is thrown in the developer log file:

Error: unexpected status code 503
    at doThing (webpack://app-package/src/issue-events.js:62:11)
    at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
    at processAttachment (webpack://app-package/src/issue-events.js:72:5)
    at Object.exports.run (webpack://app-package/src/issue-events.js:80:7)
    at async t.<computed> [as handler] (/var/task/issue-events.cjs:2:164737) 

and

upstream connect error or disconnect/reset before headers. reset reason: connection termination

Workaround

Currently there is no known workaround for this behavior. A workaround will be added here when available