This isn't a bug-report, but it's information gathering in preparation for one.
In testing the new "network-transport" communication layer, I've got a situation where killThreads hangs and I'm looking for possible explanations. It's described in this issue ticket, with some instructions for reproducing:
Because killThread amounts to throwTo of a KillThread exception, I was initially thinking that there's something stopping the receiving thread from receiving the asynchronous exception? (Like a tight, non-allocating loop?)
But the pattern of GHC versions where the bug occurs is odd (i.e. non-monotonic!). I'd appreciate knowing if anyone else sees the same results.