Actually Mark, if the kernel crashes, the socket will be closed and generate errors similar to that... Those aren't necessarily client-side problems. They're the same error codes I was getting during this ordeal (see screenshot and kernel stack trace.)

So I don't think this is a client-side problem.