Just very briefly – it seems that I managed to fix the Strathspey server memory issue by merely changing the configuration of the application server (Gunicorn) such that a Django worker process is preemptively shut down and replaced with a fresh new one after it has served 250 requests (or so – Gunicorn adds random “jitter” so that not all worker processes are cycled at the same time). The number of requests used to be unlimited, such that the worker processes would grow over time until they consumed most of the memory in the machine.

This change now means that every 5 or 10 minutes one of the worker processes will be restarted, but according to Zabbix (the monitoring tool), the memory consumption of the system is now pretty much constant at a little less than 2 gigs of RAM free on a 4-gig VM. So, sorry, Hetzner Cloud, it appears that you're not getting more business out of me for bigger Strathspey VMs.