I'm not going to enter a comment on which platform is having the problem, 
but here is an example of why it may not be NetBSD's problem.

Remember the "ping of doom"? Pings sent to a machine that were larger 
then the standard limit would cause some systems to crash. This was 
caused by the system not properly checking/handling oversized packets. 
The problem was exposed by Windows NT, but the bug involved both systems. 
NT allowed oversized packets to be sent, the recieving side failed to 
deal with it in an acceptable manor. 


Likewise, it could be that something that NetBSD is sending something 
(either valid or invalid) that exposes a bug in linux / pppd. 

I'm not trying to attack anyone here, just provide an example for better 
understanding. I hope this has been helpful.

-- Jason