Reactive programming is all about relegating the management of a program's state changes to the realm of the runtime environment. Nevertheless, sometimes it is still necessary to enrich a reactive program with state variables that are explicitly updated by the programmer. In current reactive languages this is accomplished either by polluting the reactive paradigm with imperative constructs or by relying on built-in operators such as \texttt{foldp}.

This paper introduces \emph{trampoline variables}, a new \emph{general} mechanism that allows reactive programs to manipulate state explicitly without resorting to imperative programming. We show that our proposal is at least as powerful as existing built-in reactive operators. We also analyse how reactive programs with trampoline variables can be composed and how they can form the basis to replace stateful constituents of a running reactive program — a.k.a. hotswapping — in a coherent way. The latter is an essential building block towards live IDEs for reactive programming languages.