Ideas and resources for fuzzing shared-everything threads

This issue is intended as a place for discussion about approaches we can all use to improve our assurances that we correctly implement this proposal. A place to share relevant papers, moonshot ideas, etc...

Unlike many other Wasm proposals,[^0] we can't effectively exercise this proposal's additions to the language by just simply generating pseudo-random Wasm modules that contain this proposal's new instruction and feeding those into the system under test.

[^0]: To some degree everything here applies equally to the original threads proposal, but shared-everything threads expands the scope much further.

Off the top of my head, we need to additionally do things like

• actually spawn multiple threads
• check whether the dynamic trace of accesses across threads is valid given the happens-before relationships imposed by those operations used to make those accesses
• ideally avoid dead locks in these generated concurrent programs

Start with the issue discussion and its stated constraints around spawning threads, validating dynamic traces against happens-before relationships, and avoiding deadlocks. Done would be a documented, agreed approach or set of resources for improving assurance of the shared-everything threads proposal.