I've had sporadic problems with clearing the WP cache causing the server to return 520 errors for a few minutes. Usually other sites on the same server are fine, it's specific to this vhost. Logging in via SSH, checking with htop, rsync is usually hogging most of the cpu. Restarting the fpm and then restarting Apache restores the website.
According to https://www.claudiokuenzler.com/blog/361/rsnapshot-backup-reduce-high-load-with-ionice, the big bottleneck with rsync, which rsnaphot runs on, is i/o, not cpu, and rsync can actually tie up i/o such that a web server won't respond to http requests. This can be solved by making the rsnapshot command in crontab ionice -c 3 [rsnapshot command] instead of just the rsnapshot command, which tells rsync not to wait until the disk is idle before trying to access it. So I did. In fact, I made it nice -n 19 ionice -c 3 [rsnapshot command] although according to the article this isn't likely to help much.
I think there's a problem with the caching plugin that it causes this, but hopefully reducing rsnapshot's resource consumption will solve the problem.
