It seems that every word in the TDOT Index volume is tagged with every other headword.
Confirm.
It seems to have a maximum wonkiness (probably array size), and scans the top major level, instead of the lemma milestone.
