We’re still seeing major problems with the Streaming API in the USA. At the moment we’re completely unable to use it for anything which doesn’t have very long lived connections. Are other people out there still having problems?
Something else we’ve tried here to mitigate the problem:
Previously our streamer client was stateless - so if it got a couple of 420 responses and was waiting before reconnect (60s, 120s, 240s), but then was restarted on a 5 minute interval, it would try to reconnect again immediately with those credentials - violating the proper backoff mechanism.
We’ve updated the client to maintain state when it goes into a backoff condition, ensuring that if it gets restarted on an interval, it continues to wait for the full and proper backoff period before attempting to reconnect to the API.
This has reduced the total number of 420 responses we see reduced slightly, simply because we’re not attempting to reconnect when we should be waiting. But the underlying problem seems to be just the same. So it doesn’t seem like the root cause is anything to do with us trying to reconnect when we should have been waiting, as was the behaviour of the stateless client.
