Bonus comment - this is affecting just 'that' system and I could not repro in my own CentOS 7.2
Without knowing what's going on with that system, it's going to be hard to pinpoint.
Is there anythign else you can tell me in particular about this system? Is this a "managed" AWS system??
Did somebody fiddle with the resource limits?
Was the machine rebooted prematurely before the cache was completely built?