This is the main way this feature is supposed to work:

1. you discover that a test started failing at some point
   but a misconfiguration in the CI script didn't catch that
   causing a bunch of false negative (ie false green) CIs
   over time
2. you fix the script and now want to figure out when the
   test started failing
3. you want to rebuild older commits, potentially only on specific
   targets, but with the new fixed CI script that correctly
   catches failures

This is what this commit enables