We have servers in different data centers. If multiple servers are affected, it can be some sort of backbone issues somewhere in between. Our monitoring does not detect any problems so it also looks like only certain locations are affected. We can't do anything about it really.