Once #121970 is merged and enough (all?) code is converted to HandleErrorWithContext, we can analyze logs from jobs like https://testgrid.k8s.io/sig-instrumentation-tests#kind-json-logging-master to find out:

• whether there are unhandled errors
• how many

With JSON output, jq might be enough to search for `logger=".UnhandledError.". See https://github.com/kubernetes/kubernetes/tree/master/test/integration/logs/benchmark#real-log-data for instructions on how to download those logs.

If there are none, then we might be able to add post-processing of all or at least some jobs and check that no such errors creep in. If there are some, we might need an allow list.

/cc @sttts