We could add these as a new array output field in the validate output.

The primary motivation for this would be to make remediations more simple; instead of re-implementing parts of validate in the remediation scripts (example here), we could access all the corrupted buckets and then directly run the timeseries-bucket-repair script (linked here).