I am seeing my worker processes grow incrementally, according to htop, to using more than 6% of hardware memory (16GB in this case). I have not altered the memory argument for the workers, so I believe that if a process goes over 128 MB (the default), it should die and restart. That does not appear to be the case.
Furthermore, my own tests to compare what htop reports in MEM% vs what PHP reports with
memory_get_usage(true), PHP appears to severely under-report compared to what HTOP does.
Does anyone have experience on this to help me understand the discrepencies?