I keep seeing this metric in your weekly reports for "Failed Events". If an event fails to record from a remote location via your API, how would you even know that? Do you detect that, and then send a follow-up event that indicates a failure of the original event? And if that is the case, then how likely to fail are those follow-up events themselves?
You're right that Keen is only able to track a subset of failed events. Keen sends a weekly report that shows how many events were successfully collected and how many failed (among other things).
The number of failed events refers to the requests that made it to Keen's API but still failed for some reason. Keen doesn't do any client-side tracking of failed events (for example, we don't have reporting functions in our iOS or Android libraries that tell us what is happening or failing in those clients).
Your failed event count in your report refers to events that failed for reasons like:
This is also documented in API Reference: Errors.
Your report won't include events that failed due to a dropped connection or client-side issues like not properly setting up the client.