Sorry please ignore my original question - My assumption was SG/CB was hitting an internal limit on user sessions but I was wrong. I’ll explain:
I have a setup with one writer (push repl) and many (+200k) readers (pull repls) all using the exact same account & channel. The writer is a .net c# service and all readers are iOS & Android apps.
The last few days the writer kept failing to push any doc inserts/updates. I went through the SG/CB logs and found no errors. After some more debugging I realised that the problem was the actual repl connection. For whatever reason after a certain amount of inactivity the push repl connection stops working and doesn’t recover.
This is the first time I hit this issue after working flawlessly for 2 years so I’m not sure what the root cause is. The code the pushes changes hasn’t changed at all.
As a quick fix, I’ve changed the push repl code to reconnect every so often (30 mins) to ensure updates are being pushed. That seems to work fine. When I have time I’ll debug and figure out exactly what is happening.