The thread pool starts out constant but can accumulate more threads over time. The concurrent usage of those threads seems to still respect the --jobs parameter, which suggests that some of the spawned threads may be dying.
In the ticker log, this shows up as progressively more and more lines tracking the busy jobs.
FWIW, it mostly seems to occur with parallel test execution, when there are test tasks waiting for their futures to finish, even though there are no futures remaining.
The thread pool starts out constant but can accumulate more threads over time. The concurrent usage of those threads seems to still respect the
--jobsparameter, which suggests that some of the spawned threads may be dying.In the ticker log, this shows up as progressively more and more lines tracking the busy jobs.
FWIW, it mostly seems to occur with parallel test execution, when there are test tasks waiting for their futures to finish, even though there are no futures remaining.