Potential lock up

[mstoykov]

We have received info that sometimes using this module leads to k6 not stopping the process after the test duration is over.

Unfortunately there isn't all that much concrete info. The only particular interesting thing is that there are a lot of websocket connectiongs per VU 500+.

I was thinking it might be related to

xk6-websockets/websockets/websockets.go

Lines 362 to 368 in acae12f

	case <-ctx.Done():
	// unfortunately it is really possible that k6 has been winding down the VU and the above code
	// to close the connection will just never be called as the event loop no longer executes the callbacks.
	// This ultimately needs a separate signal for when the eventloop will not execute anything after this.
	// To try to prevent leaking goroutines here we will try to figure out if we need to close the `done`
	// channel and wait a bit and close it if it isn't. This might be better off with more mutexes
	timer := time.NewTimer(time.Millisecond * 250)

but looking at the code - there is no way this code will miss to stop the execution. The problem with it, it might stop it too soon.

But it might also be some other corner case, and unfortunately it might even be in some other part that just gets trickered (more easily) by this module.

But the first step will be to make this at all reproducible I guess.