Weird spikes on multiple mysql graphs

Hi guys,

Pretty randomly we see spikes which simply don’t make any sense. For instance, IO traffic is 770 TBps or 262 Bil SELECTs. Check screenshots - it will explain everything.
We monitor multiple instances and those spikes appear for one mysql only. It’s 5.6.25-73.1-log Percona Server. MySQL’s uptime is 2+ years. For example, Com_select in SHOW GLOBAL STATUS is 1052298122417. Can it be due to large numbers?

Any ideas what’s wrong?