There was delay in sending messages in the US region and higher latency in some API endpoints. As for the latency: around 0.003% of the requests during this period had a "client disconnect" (which means that the client was disconnected before the server responded which usually means a client timeout or a networking issue), and a total of 2 "create message" calls took more than 5 seconds (across all customers, so a very tiny fraction of the requests).