The correctness fix in 811b13a causes interp to bail a lot earlier in cases, specifically leaving in many more function calls than previously. This leads to much more work for the asyncify pass (in our binary it goes from taking ~3s to taking ~300s), and the precompute-propagate pass to effectively never finish. While some of the work of fixing this relies on speeding up wasm-opt, we should also see we can be more intelligent about things on the tinygo side.