Combining parallel files leaves the old data files around.
From Dirkjan Ochtman (hg dev):
BTW, I modified our coverage.py use to use the parallel files (with the PID in the filename), but apparently the combination step leaves behind the old files (leading some people to complain about the disk use). There wasn't anything in the code that I could use for this, as far as I could see from a quick glance. Is that right? Might be a useful option for a future release.