Mild case of Stockholm Syndrom?
Yes. Whether that is good or bad is unclear. Using state is a property of the code that asks for some care. But, so is using cuts, constraints, tabling, assert/retract and a bunch more. Why would using state be so much different? Translating full DCGs (no-non-dcg code) is fairly straightforward. As we start to mix with non-dcg code we have to decide when to do the unification of the remaining tail. There is currently a fair consensus about that, but that has taken a while, introduces overhead and causes sometimes unexpected results. There has also be quite some discussion of which primitives need {}
and which not.
You have more experience with edgc, but do we not need to know how many accumulators exist? Seems edcg uses a list of accumulators. That probably gets rather slow quickly. Using a dict would allow for named accumulators.