Yeah that’d be pretty cool. But doesn’t that lead back to thread-specific signal handling which I infer from @jsbueno’s earlier post is quite firmly in the “too hard” basket at this point?
I wonder if @cmaloney has any ideas in this area? Based on this post of his in the free-threading PEP 779 discussion some of the low-level work going relating to free-threading might also open up avenues to allow aiofiles to be properly async?
Likely not given C stdlib read is blocking as mentioned earlier, but I thought it worth asking the question at least.