I've just done some trawling back through my mail archives and found the 
following interesting snippets, which taken together might be a good clue:


--------------------


>From: is@netbsd.org
Date: Sun, 1 Dec 2002 17:47:55 +0100

I'm running an 1-week-old 1.6 branch snapshot, and I'm unable to build a
working teTeX out of pkgsrc.


>From: Richard Earnshaw <rearnsha@arm.com>
Date: Sun, 01 Dec 2002 16:58:57 +0000

A kernel I built a week ago from -current is exhibiting random segfaults 
in user-land, but I haven't tried a branch kernel.


>From: Richard Earnshaw <rearnsha@arm.com>
Date: Tue, 03 Dec 2002 09:24:05 +0000

I've not seen any panics, and a shark kernel I built from Sunday's sources 
seems to be much better (at least, it can bootstrap gcc-3.3 devel without 
dieing.   I need to thrash it a bit more before I'm happy though.

------------------------

So it would appear from this that sometime just before 23 November a 
change was made and pulled up onto the branch that caused instability.  
This was fixed in -current sometime during that week but hasn't been 
pulled onto the branch.

All we need to find out now is what ;-)

R.