I hate to say this, but ... we don't. Well, I don't seem to, anyway.
Ever since I started running 1.25 in earnest (I refrained from doing so as long as I could, due to the known bugs, e.g. the RenderVolumeLOD issue that you pointed out etc.), I am getting frequent, seemingly random crashes myself. Whereas 1.23.5.36 would crash maybe once a week on me, 1.25.0.17 crashed me three times in the previous four hours alone.
The log does not indicate anything specific -- there is no indication of anything wrong or, even for that matter, a crash handler being called. The stack_trace.log contains a "0" followed by three blanks, and that's it. This is on Ubuntu 10.04.
On a (probably) unrelated note, I am observing severe client-side stalls occasionally (from sub-second to, in bad cases, several seconds) where the entire processing freezes -- no frame updates, no network communications -- and then picks up again. This does seem content dependent, I seem to be able to always observe it at the Strapped mall in the Calintreau region. Again, I can find no indication of anything amiss in the log and 1.23.5.36 does not show these issues.
Suffice it to say that currently I am not a happy camper. I am tempted to give up on 1.25 again and go back to 1.23 for as long as it can still serve me reasonably well. But obviously with all the current developments in SL, its useful lifespan will obviously be very short.
(Going off topic here: actually I also see that for 1.25 -- while I admire your excellent work, I just don't see you as a single individual being able to put in enough work quickly enough to keep up with the faster pace. And I seriously hope I am wrong.)
So, any ideas what might be happening here that's causing these (seemingly) spontaneuous aborts without crash handler log output or a decent stack trace being written? As for the stalls, could this be some odd temporary synchronisation issue between the texture decoding and the main thread of execution? Any further input on what to watch out for would be appreciated.
Love,
Lia