cc @a-siva

//cc @jonahwilliams would you be the right person to help out Derek here ?

@derekxu16 are these package:test send/receive ports? if so have you talked to any of the package:test maintainers yet? I can look at it a bit but I'm not very familiar with the work.

There was some discussion with the package:test maintainers here: https://chat.google.com/room/AAAA3-KvHNw/_Ng7rX3hbHk. Nate's last message indicated that package:test currently forcibly kills spawned isolates when they are done running test code, and he doesn't know whether or not the spawned isolates would shut down by themselves if the force-killing logic was removed. When I get the chance, I can check what happens if the force-killing logic is removed from package:test.

and he doesn't know whether or not the spawned isolates would shut down by themselves if the force-killing logic was removed.

They do in the typical case.

After some more thinking, though, I got worried about removing the kill() call. I think it might be in place as a resource optimization. Nothing breaks when I remove it - but if some test patterns are commonly holding resources which keep test isolates running don't know if it could become a noticeable performance issue.

dart-lang/test#2198 (comment)

The culprit of slow performance seems to be the fact that all dart code for all tests in one combined dill file for all concurrently-running dart isolates spawned in flutter_tester runs on a single one and the same platform thread. So effectively we end up running all the tests on a single thread, not taking advantage of possible paralellization.