This is a different proposal to fix our intermittent CI failures.

Since the tests seem to always pass when executed in isolation, this ticket proposes dividing the entire test suite into parts that can run in their own processes, thus hopefully eliminating the unknown impacts of one test on other tests that cause the intermittent failures.

The implementation takes care to always have the "default"/"remaining" bucket such that adding tests does not require explicitly assigning buckets to them.