I'm not in a position to evaluate the clang fixes, but I confirm that supernova no longer crashes when setting an unallocated buffer. (Oddly, in my environment, it takes several seconds for the error post to appear, though.)

the error reporting is done asynchronously from a low-priority thread, as writing to stdout is not exactly something that should be done from a real-time thread

Sure, that makes sense. I was surprised that it took something like 15 seconds for the messages to appear. 100-200 ms would be reasonable.