named fails answering

[Radu "Ux" D.]

Hi

Following this issue, I enabled the logs on BIND named:
logging {
  channel debugging { file "bind.log" versions 10 size 10m;severity 
debug 99;print-category yes;print-severity yes;print-time yes; };
  category default { debugging; };
  category general { debugging; };
};
The behavior stopped! named never froze again. So the logs were redundant.
After more than a week, once I removed the log configuration, named 
started to intermittently take ~25% CPU, to not answer to queries and 
updates.
The named is restarted quite often (several times per day) 
automatically. When this is happening, the service on Windows can't be 
stopped and the daemon on Solaris can be killed only with SIGKILL.

Anyone? Any idea why this is happening or how could I continue the 
investigation?

Thx, Radu
p.s. Yes, I use BIND 9.3.1 and I have to stick with it. If any issue 
fixed in 9.3.2 would affect this behavior, please let me know.
The facts are occurring on Solaris 9 and W2K also.