Synchronizing sequences are used in the context of finite state machine based testing in order to initialize an implementation to a particular state. The cubic complexity of even the fastest heuristic algorithms known in the literature to construct a synchronizing sequence can be a problem in practice. In order to scale the performance of synchronizing heuristics, some algorithmic improvements together with a parallel implementation of these heuristics are proposed in this paper. An experimental study is also presented which shows that the improved/parallel implementation can yield a considerable speedup over the sequential implementation.