"Replay detected of message" causes

[Klingenstein, Nate]

> Ok, so that's on the initial URL mainly I suppose. Is there a reason to think it wouldn't apply to the mid-conversation URLs and is there a reason to expect more cases where the original never hits the server or would you expect a replay?

It looks like about 60% replays and 40% expirations, just from a quick visual check.

> Maybe you figured out the differences wrt to webflow behavior but I could explain what it would do.

I'm still scared.

To Chris Bongaarts' message, the affected users are totally fine on the same devices on wifi networks.  It's cellular service only.  We have everything going through the same reverse NAT.

The initial reporter can't reproduce the issue today following all our testing, which makes me even more suspicious of traffic shaping being involved here.

I'm going to run reverse lookups on all the inbound addresses to see if I can trace all these back to the single mobile carrier that users have named, buuut... the web server fronting this wasn't set up to log X-Forwarded-For and we have another large production change scheduled this evening, so it'll be a couple days before I have that information.