Description:
------------
Hi there.

It is common practice to call fastcgi_finish_request as soon as response is ready to be sent and defer some heavy load tasks after that. But here is the problem: process will no longer be terminated if it exceeds request_terminate_timeout.

It happens because fpm_request_finished called inside fastcgi_finish_request changes request state to FPM_REQUEST_FINISHED. But fpm_request_check_timed_out only checks processes being in state between FPM_REQUEST_ACCEPTING and FPM_REQUEST_END.

Turns out that there is no way to limit execution time of a worker after fastcgi_finish_request. That behavior may lead to all fpm-pool being occupied.

Possible solution may be a config option which enables fpm_request_check_timed_out for processes in FPM_REQUEST_FINISHED state,

Test script:
---------------
<?php

fastcgi_finish_request();

$time = microtime(true);
// must be any time greater than configured request_terminate_timeout
$delaySec = 10;

while ($delaySec > (microtime(true) - $time)) {
	usleep(100000);
}


Expected result:
----------------
Php worker being terminated by fpm after %request_terminate_timeout% of execution time