there's a ludicrously clear analogy to v2 sandboxed process execution here (which i was saying earlier, but this hadn't materialized / wasn't announced at the last wasm meetup). my only question is should we compile python rules to wasm or should we add an interface to wasm rules to the rust scheduler (either way we lose the GIL for computational work)? that being said, since process executions are cacheable and remoteable, it may not be bad to have python code continue to be the bottleneck as it could encourage the incorporation of reusable external tools instead of reimplementing everything https://hacks.mozilla.org/2019/03/standardizing-wasi-a-webassembly-system-interface/