You need "adequate" Q, not maximum Q. You would like
the SRF of the -assembly- to be well past the operating
frequencies of interest. Bare MIM caps and bare wiring
inductance is a recipe for "mystery tank" kinds of chip
debug adventures. Degenerating the decoupling cap with
a bit of resistance (e.g. a poly under-jumper) is the done
thing. Where to put the damping resistance is a judgment
call - probably between C and supply, with C well connected
to the circuit block supplies, so the tank energy scrubs off
back into the source across the resistor. Ground connection
probably needs to be the most highly respected as it serves
both signal and power return. Making one plate of MIM be
part of the ground slab makes a lot of sense, layout area
economics -wise.