Something is causing PHP workers to hang for anywhere from 10–60 seconds. During these incidents, OpenLiteSpeed reports:

No request delivery notification has been received from LSAPI application, possible dead lock.

I have confirmed that the problem is not originating from MariaDB. The suspicion is that something in the PHP/WordPress layer is causing the workers to stop making progress, potentially a plugin.

What I need is a FOSS PHP profiler/tracer that can automatically identify the root cause — ideally showing the PHP call stack, function, source file/line, and what the process was waiting on when the problem occurred.

Current stack:

Ubuntu → OpenLiteSpeed → PHP 8.4 (LSAPI) → WordPress → MariaDB

I've already collected Linux/kernel-level evidence (D state, wchan, kernel stacks, strace, etc.), but that only tells me where the process is waiting; it doesn't identify the originating WordPress/PHP code.

What FOSS tool would you recommend for tracing this all the way back to the PHP/WordPress function responsible?