Our coverage machinery comments on a PR when it finds added lines that are not executed by any test. The idea is to help us write better tests that execute more of the new code.
However, some of the reports should be dismissed, and the natural way to do that on Github is to "resolve the conversation". Unfortunatly, on the next test run, our coverage machinery will remove all its old comments and make a round of new ones. This makes resolving old comments pointless, and also increases clutter since our own comments will stay.

A solution would be to improve our coverage machinery to keep comments that stay the same from one round of testing to the next (instead of removing and recreating them).

The challenge is to figure out if a new comment is the same as an existing old comment, even when line numbers in the diff have changed.