A simple and straightforward approach: set up a watch for each monitorized machine, look for the messages in the last n minutes and evaluate if the total is zero.
For a more complicated and escalable solution, check the watcher examples from the official repository: