Why are multiple simulations at un-needed temperatures "better" at searching the conformational space of the protein rather than say simply heating the protein when it gets stuck during a single trajectory?
"Replica-exchange molecular dynamics method for protein folding". Try reading this paper, it will definitely help you understand. In general it is easier for the system to visit a global minimum, in classical MD and usually around 300 K, which is a good temperature for biological systems, the system is usually "trapped" in a local minimum and it takes a longer simulation run to sample the conformational space (than with Replica Exchange).