Answer
'Connection reset' specifically means the underlying client socket's connection was closed by the other end while IFS Connect was still trying to read from it — caused by a network drop, a firewall, the remote application crashing, or (very commonly) an intentional close from something in the network path. If any load balancers sit anywhere between IFS Connect and the actual external endpoint (including ones on IFS's own cloud-hosting side), check their configured idle/connection timeout settings — a load balancer silently closing a connection once it's been idle or open past its configured threshold is a very common, easy-to-overlook cause of exactly this 'works fine for a while, then resets, then recovers' pattern under sustained or repeated outbound API traffic.