Thanks all. I'm now in deeper 50mm testing, in isolation (as here) or grouped. Difficult exercise, as differences are often subtle and easily affected by changing light, wind, knocking accidentally the focus ring, etc. And testing in isolation is important, and more reliable than one would think.
Getting a 50/1.8 with 0.45m minimum focussing distance (here: AI version) is definitely worthwhile, see above. Also, the pancake AIS version (ser. No. 4xxx) is definitely another optical design, and surprizingly an inferior one (lesser contrast, maybe due to the more exposed front lens; more bokeh outlining; probably less sharp but that's hard to judge given the lower contrast).
On the contrary, the 50/1.8, of which I finally got a good copy (3rd attempt) is definitely an improvement over the 50/2 AI, the most apparent being less haze and vignetting at f/2, and less nervous a bokeh. But I'll stop here: one clear finding is, also, that long- and short-distance rankings can be very different. The usual chart shooting tests deliver truncated info.
