Busy-wait loops counter-productive on slow tests

I'm using Yath to drive my Selenium tests, which are extremely slow to produce test events, when compared to the regular t/*.t test file pattern. The result is that I have one core busy-spinning the wait-loop. This core could have been put to use in producing test results faster, if Yath would fall back to select() or poll() either immediately, or after initially trying to busy-wait a few cycles.

This doesn't matter so much on my local test machine (although it would silence my laptop during test runs), but does make a difference on core-constrained environments such as GitHub Actions or CircleCI. I use both with free FOSS access.

Start by locating Yath's wait loop and inspect how it handles slow Selenium test events. Compare the current busy-wait behavior with select() or poll() and verify that slow tests no longer consume a core unnecessarily while test results continue to be produced.