Timed automata are a very successful notation for specifying and verifying real-time systems. One problem of the approach though is that timelocks can freely arise. These are counter-intuitive situations in which a specifier's description of a component automaton can inadvertently prevent time from passing beyond a certain point. This means, in fact, that the entire system stops. We identify a number of different types of timelocks and argue that each type should be treated differently. We distinguish between time-actionlocks and zeno-timelocks and argue that a constructive approach should be applied to preventing the former of these, while an analytical approach should be used to prevent the latter. In accordance with this position, we pre...