Identify source of 30 second proxy latency

[mark_hammelman]

I've recently started getting complaints from clients that proxies that normally are sub-second will occasionally take >30 seconds.

I used a custom report to help identify whether the spike is happening on the request or response, and its definitely on the 'request' side of the flow. Its always a spike of 30s.

Heres an example where you can see the request side spikes up to 30Kms while response is consistently under 200ms.

This is happening a few times a day across a handful of proxies. We don't know how to recreate it so I've never been able to catch it while running a trace.

Any advice or suggestions for what I can do to narrow it down further and identify which specific policy or sharedFlow that is creating these delays?

edit: These are executing in Edge public cloud in a paid org.