Segfault for C++ binding with Go library while cleaning up

This may be a long shot but we're out of ideas. we have a Go library with JS bindings, see https://github.com/tdewolff/minify/tree/master/bindings/js which builds a Go library (using MingW) to a C archive, then we use that together with the C file to build the NAPI binary. I use GitHub actions to build on each platform, see https://github.com/tdewolff/minify/blob/master/.github/workflows/nodejs.yml.

For some reason, executing https://github.com/tdewolff/minify/blob/master/bindings/js/test/worker.js gives a segfault right at the end (see https://github.com/tdewolff/minify/runs/7293049169?check_suite_focus=true#step:6:74) which happens while running await worker.terminate(). Is there a way to debug this segfault? It is likely a race condition since it doesn't always happen. Something probably to do with the Go garbage collector, even though turning it off doesn't fix the problem.

How can I debug this segfault? The segfault-handler package doesn't seem to support ES6 import syntax, but converting the worker.js above to a CommonJS format doesn't usually cause a segfault and if it does the handler doesn't catch it.

PS: segfault happens on Ubuntu/Arch, but not on MacOS. For Windows I run into another problem

Start with bindings/js/test/worker.js and the .github/workflows/nodejs.yml workflow, then inspect how the Go library is built into a C archive and used by the N-API binary. Reproduce the cleanup crash on Ubuntu or Arch and compare behavior across platforms. Done requires identifying the cause of the intermittent segfault and documenting or implementing a verified resolution.