Abnormally slow requests on static mirror hosts

This morning Nagios was unhappy with some of the static mirror hosts, with several errors like this:

tor-nagios: [web-chi-03] network service - https is CRITICAL: CRITICAL - Socket timeout after 10 seconds
tor-nagios: [global] mirror sync - www is CRITICAL: CRITICAL: 38.229.82.25 broken: 500 Cant connect to www.torproject.org:443

Looking at Grafana, since about one week ago we are seeing increased loads on our web mirrors, which Apache connection slots getting abnormally filled up:

Currently Nagios only barks if the web hosts don't respond to HTTPS connections within 10 seconds, which is fine to the purposes of determining whether the service is alive at all, but for static sites even on a busy webserver response times of 1 second or more shouldn't be considered acceptable.