First, SQS considers sends, receives and deletes as requests. So that's
>>> 600000 / 3
200000
tasks.
Last month there were 31 days. So with the check for events task alone there were
>>> 60 * 24 * 31
44640
tasks or
>>> 44640 * 3
133920
requests used upfront.
Now, you'd only have to average
>>> (200000.0 - 44640) / 44640
3.4802867383512543
events per check_for_events invocation to get to 600k requests.
Now I'm not certain what volume of events you're dealing with, but perhaps this puts the usage more into perspective.