I'm using stretch to lay out a tree-like structure. With 63 nodes and a maximum depth of 16, compute_layout takes about 5 seconds to complete. Profiling illustrates that compute time increases exponentially with depth, and that compute_internal is called over 50k times for my structure of 63 nodes.

Is this sort of performance expected? Is nesting this deeply just an entirely unexpected thing to be doing with stretch? Are there constraints I could be putting in place to elide some of this recursion?