One of the problems I noted last time we went around on this is that
(assuming a limited amount of memory is available, of the order of
100Mb) the recursion can fill up available memory sufficiently quickly
that your VM is actually _dead_ by the time you press the interrupt
key(s). It spends a long time attempting to print out the
lowspacedebug.log files before actually going castors up and yo
unotice. Thus, you can spend a great deal of frustrating time trying to
interrupt whilst the system is deceased, no more, gone to meet it's
makefile etc etc.

IIRC we also noticed that the lowspace interrupt response had some
issues (which is how come the recursion could fill up memory) and a fix
was produced for that. There ought not be a problem anymore.

tim
--
Tim Rowledge, tim at sumeru.stanford.edu, http://sumeru.stanford.edu/tim
42.7 percent of all statistics are made up on the spot.