CI: Large spike in scheduling failures with the CapacityFilter since March 13

Scheduling failures with the CapacityFilter have spiked since March 13th:

http://logstash.openstack.org/#dashboard/file/logstash.json?query=message%3A%5C%22Filtering%20removed%20all%20hosts%20for%20the%20request%20with%20volume%20ID%5C%22%20AND%20message%3A%5C%22CapacityFilter%3A%20(start%3A%201%2C%20end%3A%200)%5C%22%20AND%20tags%3A%5C%22screen-c-sch.txt%5C%22%20AND%20voting%3A1&from=7d

994 hits in 7 days, all failures, check and gate queues.

This is impacting the integrated gate jobs and our ability to merge any code.

Were there any recent changes in the last few days to tip this over? Maybe new tests which are now running concurrently and consuming more space on the host?