Comp-head phrase clashes with its parents; question particles

I have a grammar for which the choices file specifies SVO word order (so, a head-comp phrase) as well as sentence-final question particles (so, a comp-head phrase).

The grammar doesn’t load and I don’t quite understand why. The error message is that the comp-head phrase clashes with its parents, and removing the HEAD comp constraint fixes that, but why is there a clash in the first place? None of the phrases parents say anything about HEAD, as far as I can see? I include both the addenda in mylang.tdl and the matrix types:

; comp-head-phrase requires things that are [ HEAD comp ].

comp-head-phrase := basic-head-1st-comp-phrase & head-final &
  [ SYNSEM.LOCAL.CAT.HEAD comp,
    HEAD-DTR.SYNSEM.LOCAL.CAT.HEAD +nv ].

basic-head-comp-phrase :+ [ SYNSEM [ LIGHT #light,
             LOCAL.CAT [ HC-LIGHT #light,
                         MC #mc ] ],
    HEAD-DTR.SYNSEM.LOCAL.CAT.MC #mc ].


basic-head-1st-comp-phrase := basic-head-comp-phrase &
  [ SYNSEM.LOCAL.CAT.VAL.COMPS #comps,
    HEAD-DTR.SYNSEM.LOCAL.CAT.VAL.COMPS < #synsem . #comps >,
    NON-HEAD-DTR.SYNSEM #synsem ].

basic-head-comp-phrase := binary-nonloc-phrase & head-compositional &
              binary-headed-phrase &
  [ SYNSEM phr-synsem-min &
           [ LOCAL.CAT [ WH [ OR < #or1, #or2 > ],
                         VAL [ SUBJ #subj,
                               SPEC #spec,
                               SPR #spr ],
                         POSTHEAD #ph,
                         HC-LIGHT #light ],
             MODIFIED #modified ],
    HEAD-DTR.SYNSEM [ LOCAL.CAT [ WH #or1,
                                  VAL [ SUBJ #subj,
                                        SPEC #spec,
                                        SPR #spr ],
                                  HC-LIGHT #light,
                                  POSTHEAD #ph ],
                      MODIFIED #modified ],
    NON-HEAD-DTR.SYNSEM canonical-synsem & [ LOCAL.CAT.WH #or2 ],
    C-CONT [ RELS.LIST < >, HCONS.LIST < >, ICONS.LIST < > ] ].

Does anyone have an idea? MC and LIGHT and HC-LIGHT are appropriate for any CAT…

For some reason, this rule is specifying HEAD values for both mother & head daughter, and they are incompatible:

comp-head-phrase := basic-head-1st-comp-phrase & head-final &
[ SYNSEM.LOCAL.CAT.HEAD comp,
HEAD-DTR.SYNSEM.LOCAL.CAT.HEAD +nv ].

(The conflict with the parent is presumably with head-compositional, which identifies those HEAD values.)

So, if a language has clause-final particles but subordinator-complementizers which go before the causal complement, that would not be possible to model via these HEAD constraints, would it? Would need other features (INIT?)

Either different HEAD types for those, or an INIT feature. But INIT should work, because these are all things that only pick up one complement, right?