Yeah it seems to, even performing no work in there actually (just logging the change to console) seems to eventually disconnect the watch client if enough changes are created to target resource in quick enough succession. I think the one caveat to this is that we're using a self-generated .NET client, reckon that might be the cause?