I'm not recommending either way, but believe that
managing a large panel by analog signals (which in
the end will still be made digitally) will be much more
of a noise sensitive wireball. That in fact was the old
way. Layers behind the panel with the phase / amplitude
stuff and MxN independent feed wires to the MxN panels
(or maybe 2xMxN if the TX and RX go by different wires
to / from the on-element circulators).
Nowadays a commutated RF switch may replace the
circulator, and local gain / phase control lets elements
share a common feed and serial bus for great savings
in copper and structural weight.
Switches and passives will be much more stable in
hostile environments, than a "Vtune for gain and
Vtune for phase". Analog "might" do better for fine
resolution, but if that ends up "in the noise" for
"that sure is a big ball of unintended antennae"
then digital stays clean a lot better. Especially if you
keep a discipline of not chattering a bunch of serial
data during time-of-flight-and-return.
I'm not a radar guy but I've done 4 element-control chips
for space radars, over the years. Now you know what I
know, about the bigger picture.