Hi, Do you know your per second and minute burst rates? Peak hours, from all our deployments, I'm showing it can go as high as 11.3m requests per minute towards api.push.apple.com. Your burst rates could be overtaxing your host setup perhaps. Well the issue is happening in vastly different parts of the world, with independent networking and hardware. Some areas are off-peak and some are peak hours, despite this the issue seems extremely coordinated every time (down to a specific minute). For example large incident was on 14th March around 10:21 UTC: Is it the outgoing connection and sending taking too long, or the response is not arriving in time. If you can provide apns-ids of a few such requests which happened in the past few days, we can check if a specific issue is visible on our end. I'll do some digging to be able to answer these (thought due to the nature of the errors we are seeing, I expect it's timeouts because response isn't received ie. request headers and body are uploaded but then no response frames ever arrive), it might take some time though. We don't log the apns-id but maybe it's time to start and add it to the code as it might be useful in the future for other investigations as well.