debugging advice needed

I'd like to ask you to point me what to do in the following case: the software is: Solaris 2.4, Apache 1.3.12, MySQL 3.22.25 and PHP php4-200010221045 snapshot. Sometimes I can observe several httpd processes looping. They eat 100% CPU. What I did was: I run truss -topen,read,write,close on main Apache process, saving the output to disk. Seeing Apache process looping, I killed it -9 and examined truss output for the looping child process (grep looping pid). I found this httpd process was serving dynamic page content from PHP+MySQL page. Last read was a read from MySQL server (select query), and it seems to me it was successful (the row could be seen in truss output). The last lines of truss output were: 21266: Received signal #13, SIGPIPE [ignored] 21266: Received signal #29, SIGPROF [caught] The problem is, the page (BTW with output buffering) is in principle (almost?) always served correctly. It means I can't observe it is truncated. Only sometimes I see httpd loop. The question is, what would be the methodology to find out what part of the software doesn't work correctly? I don't suspect PHP page. I don't know what is the reason of PIPE and PROF signals - is it due to kill -9 httpdpid, or a part of software doesn't function correctly? Maciek