TestAntsPool hangs when the package is run with -count>1

TestAntsPool blocks indefinitely when the test binary runs it more than once in the same process, so go test -count=2 or higher never completes and the whole package times out.

Reproduction

$ git clone https://github.com/panjf2000/ants && cd ants
$ go test -run 'TestAntsPool$' -count=2 .
panic: test timed out after 10m0s
	running tests:
		TestAntsPool (7m40s)

-count=1 is unaffected.

What I measured

Thirty-five isolated attempts at 107e3767, each a fresh process, with -shuffle on and a four-minute ceiling:

Deterministic — a single extra iteration is enough.

Goroutine state at the timeout

testing.(*T).Run is waiting, while janitor goroutines from several pool instances are still live:

goroutine 1 [chan receive, 7 minutes]:
testing.(*T).Run(...)

goroutine 246 [select]:
github.com/panjf2000/ants/v2.(*poolCommon).purgeStaleWorkers(0x140000a8780)
	ants.go:261
created by github.com/panjf2000/ants/v2.(*poolCommon).goPurge

goroutine 247 [select]:
github.com/panjf2000/ants/v2.(*poolCommon).ticktock(0x140000a8780)
	ants.go:304
created by github.com/panjf2000/ants/v2.(*poolCommon).goTicktock

I have not diagnosed the root cause and I am not claiming a deadlock in the pool itself — the blocking may well be in the test rather than the library.

Why it might be worth fixing

go test -count=N is the usual way to surface flaky tests, and -count with -shuffle is what catches failures that depend on ordering or on state left behind by earlier tests. At present that technique cannot be used on this package at all.

Disclosure

I found this while testing flakestat, a flaky-test detector I wrote, against real Go projects. ants was picked because it is actively maintained and concurrency-heavy, not because I suspected anything. Happy to run further experiments if any would help; I did not open a PR because I do not know whether the right fix belongs in the test or the pool.

Environment: go1.25.4, darwin/arm64, commit 107e37678122b00e1b365636ce4f80623ba41579.