I finally have some real progress, with hope to improve the look for 2016 to more or less acceptable level. The best result came from flattening a few key tiles in level 17 in tooledit, and then flattening a rather small area in .flt file:
; flat 3 km radius circle at Baikonur-2
ELLIPSE 107 +63.3176154 +45.9151219 3000
; minimum pad 1 dugout
RECT 70 +63.343 +45.9193 150 50
Just a small 3 km circle centered at Site 2 flattens 5 or 6 most important areas, including Pad 1. With converted surface textures, this produces quite nice result even without pits. Some texture work is still needed - dynamic texture look with changing zoom is still quite uneven and jumpy.
Digging pits, however, is still a long shot. Test dig finally looks promising, and shows that the task can indeed be achieved... in principle:
Even at the minimum possible intervention at a wide base, shadows get still distorted even at perimeter distances (red lines). Guess this is because fence runs close to the pit right close to its placing point. Even though I got pit as small as 150 by 50, I am still having too inclined "ramps", only about 30 degrees:
This leads to... well, we already discussed that base coordinates should be moved away from the pit to some "neutral" place. Far from ideal (base coordinates are selected for a reason), but in my case will lead to moving hundreds of objects (each base has up to 5 partially repeating variations). Though this can be scripted, the next thing can't: as the dig gets under the centre of the pad, each object within it ends up either on pit or on a slope. Incidentally, this is the most built-up area, so the dig will affect elevation of way too many objects of the base. And those will have to be adjusted manually. In up to five configurations. Only for 2016 (looks like I still can't do a full switchover within at least another year).
So, to summarize: the trench-digging approach may already be a great option for new development. But still not quite suitable for converting existing intensely-populated bases. So, I'll probably exclude pitting from my current iteration of 2016 improvements, and delay a shot for a full switchover. This should still get me ~70% of what is needed for that - not bad at all. Thanks a lot for all your help, much appreciated!