Investigate performance timeouts of Repository Tree API on .com
Monitoring on .com has detected that the Repository Tree API is regularly failing with timeouts (>30s) (link).
This is curious as the endpoint in our own performance testing is typically very performant and returns around the 100ms mark (TTFB P90). Examining the calls in some detail shows that they are being called with an increased per_page
count of 100 compared to the default of 20, in a consistent manner suggesting some form of automation or tool, but even testing with this change internally doesn't cause the problem.
This may be only a problem on .com specifically that only happens sporadically. Even manually testing the same calls that reported as timed out in the monitoring has the endpoint returning in normal times.
Issue is to investigate what's happening here on .com to identify if it's a problem with the service or with the endpoint itself.