False alert issued after recovery

Tags: #<Tag:0x00007f84a6dfa3b8>

Has anyone seen this behavior: An recovery alert is issued (see screenshot), and the additional “false” alerts are issued a few hours later. I can confirm (see screenshot of graph) that the SWAP usage has not increased past the alert threshold.

I’ve marked the times in the graph to show that a recovery alert was issued, and then four false alerts were issued.

Does anyone have any tips on how to debug this?