x/net/http2: apparent deadlock in TestTransportResponseHeaderTimeout_Body on windows-386-2008

[bcmills]

panic: test timed out after 10m0s
…
goroutine 10660 [semacquire, 9 minutes]:
sync.runtime_Semacquire(0x13397808)
	C:/workdir/go/src/runtime/sema.go:56 +0x36
sync.(*WaitGroup).Wait(0x13397800)
	C:/workdir/go/src/sync/waitgroup.go:136 +0x87
golang.org/x/net/http2.(*clientTester).run(0x130e9110)
	C:/workdir/gopath/src/golang.org/x/net/http2/transport_test.go:860 +0x190
golang.org/x/net/http2.testTransportResponseHeaderTimeout(0x130ea0f0, 0x1)
	C:/workdir/gopath/src/golang.org/x/net/http2/transport_test.go:2216 +0x151
golang.org/x/net/http2.TestTransportResponseHeaderTimeout_Body(0x130ea0f0)
	C:/workdir/gopath/src/golang.org/x/net/http2/transport_test.go:2164 +0x26
testing.tRunner(0x130ea0f0, 0x823e98)
	C:/workdir/go/src/testing/testing.go:1449 +0x113
created by testing.(*T).Run
	C:/workdir/go/src/testing/testing.go:1496 +0x374

greplogs -l -e 'panic: test timed out after (?:.*\n)*golang\.org/x/net/http2\.TestTransportResponseHeaderTimeout' --since=2021-01-01
2022-05-01T00:24:25-2871e0c-d09ca2c/windows-386-2008

(CC @neild @tombergan)

[bcmills]

Since we've only seen one of these so far, I'm leaving this on the backlog in order to collect more data — it isn't clear to me whether this failure mode is specific to Windows (in my estimation somewhat likely), specific to 386 (fairly unlikely), or completely platform-dependent (fairly likely).

[Jorropo]

If it's new, probably related to one of go-review.googlesource.com/c/go/+/403136's bumped commits.

[bcmills]

@Jorropo, this failure was observed on the x/net builder, so the version of x/net vendored into the main repo should not be relevant. 😅