We get an http 500 as follow :
When facet.missing is turned on and combined with facet.pivot (if one of the pivot-faceted fields have missing counts -)
Ideally, the decission tree could be computing for the missing "entries" using the
query but it might be a performance issue on a large index (I guess)
The fallback to this could be to raise a 400 error with a clean message telling that both parameters can't be combined and then the documentation should be modified accordingly.