I've seen a few intermittent failures occur where elements are in the list of filter items when they shouldn't be. Due to the random nature of the test, it's hard to narrow down these cases. The random set that the test produces isn't random enough to cover a lot of cases but is random enough to cause intermittent unexpected behavior that is difficult to track.

I propose we refactor the workspace filtering tests to be explicit on a case-by-case basis to remove randomness. If we believe fuzzing is important for these tests, we should extract the filtering logic outside ofcoderd to a new package that doesn't requirecoderd setup.