Comment 3 for bug 541316

We are seeing this same issue at the APS with versions of the gateway built on both R3.14.8.2 and R3.14.10. In this case the client subscribes to alarm monitors for large number of channels and then basically just sits waiting in ca_pend_event(). We've had error messages logged by the gateway after midnight when there was definitely nobody starting or stopping the client (so it's not actively clearing channels), and on the client it looks like we're getting alarm value changes reported, although it's harder for me to be sure exactly what is happening there.

It's not obvious to me whether this is a server or a client library problem. If it helps we might be able to attach wireshark to the TCP circuit to track down what's actually happening.