E-matching is the problem of finding all matches of a pattern in an e-graph. Relational e-matching, a recent approach inspired by the databases community, has found succcess in equality saturation engines such as egglog. However, until now, this technique has not been implemented in a Satisfiability Modulo Theories (SMT) solver. The key difficulty of implementing relational e-matching in an SMT solver is \emph{backtracking}, i.e., the ability to undo previous matching decisions. We provide an initial implementation of a backtrackable, relational e-matching algorithm in Sundance, an SMT solver currently in development. We describe our approach of maintaining backtrackable indices and our e-matching algorithm based on a left-deep binary join. We evaluate our implementation on a set of synthetic benchmarks and show a median speedup of $25.0 \times$ over a top-down e-matching algorithm. However, when considering a set of benchmarks from the program verifier Verus, we see a median slowdown of $1.77 \times$. We conclude with a discussion of future work to close this gap and make relational e-matching a practical component of an SMT solver.