we are generating almost-correct trees. two issues remain: how to handle the starting term (rn our approach causes a weird duplicate) and also understand why we are getting anomalously permuted order within a level of the tree

convert tokens to name (with an awful ad-hoc scheme that we should revise later) to demonstrate end-to-end parse correctness

prepared instrumentation for tracing the mutual recursions to investigate this bug (which looks suspiciously like a highly elaborate off-by-one)