Too many jobs/threads with parallel testing

[lefou]

I tried to run some tests in the current main branch as of 15924c9 and notices some way to high job count.

> cat .mill-opts
--jobs=0.5C

> mill __.local.server.testCached
...
[15282/15293] ============================== __.local.server.testCached ============================== 190s
[14966] integration.invalidation[process-file-deleted-exit].local.server.testCached 39s
[14967] integration.invalidation[version-change].local.server.testCached 39s
[14967-0] worker-0 39s mill.integration.VersionChangeTests
[14968] integration.invalidation[zinc-build-compilation].local.server.testCached 39s
[14968-0] worker-0 39s mill.integration.ZincBuildCompilationTests
[14969] integration.invalidation[invalidation].local.server.testCached 39s
[14969-0] worker-0 39s mill.integration.ScriptsInvalidationTests
[14970] integration.invalidation[codesig-subfolder].local.server.testCached 39s
[14970-0] worker-0 39s mill.integration.CodeSigSubfolderRenamedReorderTests
[14971] integration.invalidation[run-background].local.server.testCached 39s
[14971-0] worker-0 39s mill.integration.RunBackgroundTests
[14972] integration.invalidation[selective-execution].local.server.testCached 39s
[14972-0] worker-0 39s mill.integration.SelectiveExecutionChangedInputsTests
[14972-1] worker-1 39s mill.integration.SelectiveExecutionChangedCodeTests
[14973] integration.invalidation[codesig-hello].local.server.testCached 39s
... and 310 more threads

The ... and 310 more thread is way above the expected. I have 16 logical cpu cores. Given some blocked/parked jobs thread, I'd expect a total number below 20 jobs.

[lefou]

Same, when I use --jobs 8 explicitly.

[lihaoyi]

I think this is a duplicate of #4611

[lefou]

Fixed in #4702