At the moment all thorslaves use a common temp directory (for internal spills etc). The temp file names themselves are based on PID, so there's no clash, however, then a Thor recycles, it clears the temp directory of these temporary files at startup.
With all sharing the same directory and no ready way to identify which came from which slave, every slave tries to delete all of them.
Whilst this doesn't in itself break anything, it can cause a lengthy delay as it cycles through potentially a lot of small spill files.

The easiest fix is to create a temp directory based on the slave port number.