What you call "shielding" may in fact be unacceptable clock
branch loading. Cut radiated emissions but push more current
spikes into the power/gnd planes (or more likely, the spindly
little busses you see in Big Digital cell libraries).
Maybe you'd be better off routing complementary clock pair
and criss-cross it every so often so that "far" field common
more emission is first-order nulled.
You could make a fully enclosed transmission line structure
if you cared to burn the layers and area. But now your clock
load impedance is very low, and stubs may still emit at a
significant level.
If you are worried (say) about contaminating the "little RF"
with the "big D" there's nothing like distance (physical, and
electrical) to help with that. Look to the current loops and
how longitudinal currents may cause magnetic, as well as
EM, coupling.
It's black art, if there was an easy answer it'd be well known.