We've seen a lot of flakiness because of pop-ups coming from 3rd party sources. Taking a screenshot can help debug the issue without needing a debug VM and reproducing the flakiness conditions.

This could be performed by qt-testrunner.py when a testcase re-run fails repeatedly.

Each OS should be handled differently. Feedback is needed on how to handle each platform.

EDIT: a full screen recording of a VM's screen is also a possibility, although it might generate too many video files in the end.
We could limit the storage that all videos would need by keeping only videos of workitems that failed, and delete after a few days.