Description
While fixing and issue in HistoryIntegrationTests#testThatHistoryContainsStatus in #38505 it became clear that the initial issue wasn't reproducible using just the random test seed because it uses some current System time via the ClockMock. Ideally we should be able to reproduce issues by supplying just the test seed when re-running the test.
I'd like open this issue to investigate ideas and the feasability to improve this, maybe only for HistoryIntegrationTests but also maybe more generally for other Watcher or time-dependent tests.
Maybe this isn't possible for HistoryIntegrationTests because the nature of the things tested require some sort of advancing clock, but maybe parts of it can be controlled by a reproducible random source. See #38505 (comment) for some more context.