I therefore propose that we make the RTS save & restore the value of errno in the TSO, thus making errno "Haskell-Thread-Local-State", and solving all of the above problems.

Actually this has been on my ToDo list for a long time, I just never got around to doing it (it has been mentioned on the FFI list before, IIRC).

I also propose adding a note to the FFI spec (or the bound threads document, I'm not sure which is the right place) that says that implementations have to make sure that "Errno" is local to each *Haskell* thread.

Not the FFI spec - it assumes a single-threaded Haskell implementation. The right place is the bound threads doc, and eventually the Concurrent Haskell addendum. Cheers, Simon