In respect to Consists/Lashups the MS2 by itself (i.e. talking direct to a TrackBox (6011x) out to the track and not as an auxiliary throttle connected to a Central Station) has no such M.U. function.

Nor does mFX (protocol) offer a M.U. address.
However the mFX decoder may be set to DCC (disable mFX), whereby you may make use of the "DCC M.U. address setting"

The bottom level of consisting would be to use either MM/Fx or DCC and set each loco to the same address
(this is usually only good if you never want to break up the consist or use distinct function settings)

As to load compensation, you do have to pre-match the driving characteristics of each loco "across the spectrum" of speed steps AND any Acceleration/Deceleration settings (Also shunting mode % if used).

With 3 locos of the same type (gearing/decoder) this matching is pretty well done for you.
If you are happy to "drive" your loco speed, then disabling inertia is an option - with the possible loss of some sound characteristics

From there, reality takes over with the locos taking the load as they would in the real world.
I cannot say that I have experience with a pusher/banker as you describe, and I might be inclined to adjust the characteristics of the trailing loco to be a little slower so that it isn't too pushy - or to be more precise, doesn't push too soon.
Also, as long as you do not use "dead section" , where a single loco looses power all together, the multiple heading of locos with modern decoders works very well

Wrong (yet again). There are specific CVs to set what sounds respond to the MU address, or don't respond, as required. Using these CVs you can disable the whistle/bell from resonding to the MU address, but it will still respond to the non-MU address. Equally you can have the motor sounds respond to both the MU and non-MU address, as desired.

its not easier if i only have MS2