What have you looked into on your end?

For example, do these disconnects line up to a deployment?

they do not

the disconnects are ALL devices connected via WS. its 15 devices or so at once all disconnect at the same time.

Do you have timestamps?

Yes

Apr 21, 1:01:11.250 PM UTC was the most recent

Ive looked into the WS implementation and the REDIS cache. Nothing is pointing to my set up (but obviously it could be......) just not sure what to look into next.

How long was the connection held open prior to that disconnect?

2026-04-21T10:30:27.875944Z

About 3 hours.

Ive had these WS stay open for days however in the past few days its shorter sessions

What is an "edge device" in this context?

Its a computer on a truck

Then I could only assume it's because we cycle the TCP proxy at times, and i'm sure Fastly does too.

Is fastly who Railway uses?

Fastly is the CDN provider.

Okay. What do you think about it not happening as frequently as it is now?

Its every 3-4 hours or so

I think that's fine. We don't guarantee any timeframe for how long you can keep a WebSocket connection open.