Currently, roachtest logs the error from github.MaybePost [1], but doesn't actually fail the run. In CI, this is a problem because a test failure may go unreported. Based on my preliminary check, of test.log of several dozen recent failures, there are no occurrences; i.e., it's rare, but we know that GH outages do happen from time to time.

In [2], we added logging to print the issue that was created or appended. Yet, we're missing an active alerting mechanism to detect "silent" issuePoster failures. Failing the run would also fail the CI job which is actively monitored by testeng.

[1]

Jira issue: CRDB-50886