(In reply to comment #12)
> Created attachment 5052[details]
> Avoid logging and any complex operations in a sigchld handler
> Would something like the attached patch still do the job
> while avoiding the problem?
> (My brief testing under perl 5.14.2 looks alright even when prefork
> is adjusting the number of child processes, but I'm not running spamd
> in production, so more stress testing is appreciated).
It's been a few weeks so perhaps commit to trunk and let's get more production