One of the ambiguities is that we have [ucmra_cmraR; constRF; rFunctor_diag] and [constURF; urFunctor_diag] both go from ucmraT to Funclass.
@jjourdan@robbertkrebbers is there any trick to fix such ambiguities when you have diamonds in your hierarchy? Can we reasonable make one of the involved coercions not a coercion?
Okay so looks like it's not a Coq bug -- we have had these ambiguous paths all along, but so far that would not show up as warnings when using coqc.
According to https://github.com/coq/coq/pull/9743, ambiguous paths warnings are not shown when things are convertible, so that might also be a way forward.