There is a web server, Apache, which all of a sudden started to die in a funny way. It did not die completely, just did not serve pages… Error log was littered with “zend_mm_heap corrupted” and “child pid nnnnn exit signal Segmentation fault (11)” entries and after a while it stopped serving pages (I guess all threads went zombies or something). The problem seems to lay with the Zend framework, many people reported the same (?) problem. But either developers can’t reproduce it, or they did not have a chance to fix it yet, the problem is still there. Some people cite the blessing effects of USE_ZEND_ALLOC environment variable. I added it to /etc/apache2/envvars (Ubuntu server) and the problem went away after a restart. Some advise to adjust the PHP output_buffer variable as well. I had that set to 4096 when I saw the problem, setting it to 8192 did not improve the situation. I think this is not the solution.