i'm wondering what's happening if you leave it on... things should work properly with it on but if not, we (TINW) need to figure out why...
There's two observations which are possibly relevant:
1) I've noticed all sorts of random, non-reproducible glitches (SRB sep was and the boosters overtook the Shuttle, the AP at a transition went crazy, the sound stopped working, the launch monitoring routine referenced a property which it conceptually should not be able to and crashed...) - and that smells like a general memory corruption issue. These issues pretty much disappeared when switching the new Nasal GC off
2) eatdirt specifically has noted segfaults when trajectory predictions are computed. Since trajectory prediction uses thread.newthread() to not block the main loop, chances are that the threaded GC does not take into account the possibility that Nasal itself may run in its own thread and that occasionally causes a segfault - again these crashes did not occur when testing for 24 hours with the threaded Nasal GC off
I believe there's also been a backtrace posted on the mailing list for a #2 crash