[mythtv] Mythtv crash with "incorrect codes" Again

[Ray]

I posted about this problem a while back however after some hardware testing
(and nothing else that I can think of) the frequency of the problem went way
down to the point where I thought it was gone.  Unfortunately it's still
with me though but now it only happens after approximately 8 hours of record
time not necessarily in a row.  

My problem is that sometimes while watching live tv or a recording Myth will
suddenly exit with the error "incorrect codes".  As Matt mentioned this is
actually coming from avilib rather than Myth.  I've also determined that the
problem is in (or at least triggered by) the recorded file since if I
restart Myth and play the same recording again it fails at the same spot. 
If I play the recording with mplayer (and the patches posted to this list) I
get the same failure in the same spot.  If I fast forwared past the spot
where the error would occur then I can continue playing with no problem. 
I've since switched back to using my older Myth computer (dual Celerons @
500Mhz)  and am still having the same problems with roughly the same
frequency so I think I can safely rule out a hardware problem.  I think
troubleshooting this by taking Myth out of the loop and recording with other
software is going to be difficult since the problem occurs so infrequently
(although I will try).

If you havn't given up reading the above ramblings I could really use some
suggestions:

1.  Any general suggestions on how to track this problem?
2.  Any ideas of how I could detect these bugs in the recordings short of
actually watching the whole thing?

Also specifically to Isaac and other Myth developers, would it be possible
to make Myth more forgiving of the occational error?  It seems to me that
with the combination of huge file sizes, heavy system loads, and often beta
capture drivers (not to mention audio) there is a good chance that the
occasional file error is going to happen to most users eventually and it
would be nice if Myth didn't go completely belly up when it did.   
 
Some things I'm planning to try:

1.  Different compiler, I'm currently using gcc 2.95.
2.  Different capture driver/options/compiler
3.  Look at the code for mjpeg spport in avilib, maybe I can make avilib
handle the problem frame(s) better.
4.  Creating some different load patterns on the system, generating
some heavy disk and cpu loads while recording to try and get bug to show
it'self on command.

Any advice much appreciated.

-- 
Ray