I experienced a lovely issue last week where all of my computers ktchecked at their scheduled time. In the middle of the ktcheck sessions, NetInfo/lookupd on my radmind server broke. As a result, the computers hung indefinitely at the ktcheck stage. I would expect that there would be some sort of timeout, and the connection would be broken off, either on client end or server end.
I imagine the problem exists at one of two junctures:
1. ktcheck can successfully connect to the radmind server, and has no limit for a mid-connection timeout
2. The radmind server binary holds the connection open while it is trying to access a file, or do some event controlled by lookupd. With no timeout there, it holds indefinitely.
I'm not sure which is more likely the culprit since i don't know the back-end code, however I do know that even after rebooting the radmind server (effectively killing any and all TCP sessions), the clients' ktcheck remained in a hung state (one of which was hung for a week and a half before I found it).
I don't expect that this is a terribly common scenario, and the cause of this specific scenario should be eliminated in 10.5. However, I can imagine it would probably be best for there to be some sort of timeout in situations such as this, in the event that they do arise in the future.