Worker terminates with error in Mac OS

### Version 16.16.0 ### Platform macOS 11.6.8 5 20G730 ### Subsystem _No response_ ### What steps will reproduce the bug? Examples in our CI: - https://github.com/vitest-dev/vitest/runs/7704223380?check_suite_focus=true - https://github.com/vitest-dev/vitest/runs/7685754161?check_suite_focus=true After all tests are done, we are getting the error: ``` Assertion failed: (napi_create_external(env, fseenv, fse_environment_destroy, NULL, &result) == napi_ok), function fse_environment_create, file ../src/fsevents.c, line 87 ``` I wasn't able to find this file in Node repo. We are also getting `Empty MaybeLocal` error sometimes, which _might_ be connected: - https://github.com/vitest-dev/vitest/runs/7685755871?check_suite_focus=true - https://github.com/vitest-dev/vitest/runs/7685757762?check_suite_focus=true I am not able to reproduce it on smaller scale for now, unfortunately. It seems, there is some kind of race condition. Code that fires up workers: - https://github.com/vitest-dev/vitest/blob/95c1d34368d799da417da219d394545ee83164e2/packages/vitest/src/node/pool.ts#L96 You can also notice that errors happen with "close with time out", this code is triggered here: - https://github.com/vitest-dev/vitest/blob/95c1d34368d799da417da219d394545ee83164e2/packages/vitest/src/node/core.ts#L436 ### I am able to reproduce it almost consistently in Vitest core tests: - Download [Vitest](https://github.com/vitest-dev/vitest) repo (only `node/reproduce` branch is needed) - run ``` pnpm install cd test/core ppm run test ``` If you comment out `this.pool?.close()` (which is an async noop), you should see error less time. But I got a new error with it: ``` node[28013]: ../src/node_file-inl.h:160:virtual node::fs::FSReqPromise<node::AliasedBufferBase<double, v8::Float64Array, void> >::~FSReqPromise() [AliasedBufferT = node::AliasedBufferBase<double, v8::Float64Array, void>]: Assertion `finished_' failed. 1: 0x102dd46e8 node::Abort() [/<root>/.nvm/versions/node/v16.16.0/bin/node] 2:…