Ultimately it depends on the music-pianist-piano-hall combination.
I agree 100% with your view: they have been building piano's for about 200 years, to sound good 90deg sideways, at a distance, and with the audience often sitting lower than the soundboard.
We are lucky to have one 'magical' chamber music/piano solo hall (concert hall of the Royal Conservatory in Brussels). It is a mid size shoebox (660 seats), with the back of the hall curved, probably a bit higher than it is wide. All wooden construction, with a glass ceiling (daylight !). It was constructed around 1870-80.
Whatever we do, wherever we go, there is no way to get a piano sound as good as here (even when the instrument itself is in fact superior): in this hall the 'audience' approach works great. The mic will be on axis, from 1m to 3-4m from the instrument, distance and height depending on the music and pianist. (and directivity of the M mic - yes we always do this in MS). On axis means the mic is at 90deg to the piano, and aimed almost exactly at where the curb starts at the right hand side. Height can be as low as the soundboard to 2.5m up.
But, then we get into different conditions, eg a large chamber, with 1/2 Bechstein, built in 1910. All of a sudden, there is a mic inside the instrument, and a AB couple 2m out. In this case it was the most convincing piano sound.
The point is, a piano is big. It radiates sound in all directions. Some engineers solve this by putting mics all around the instrument, which seems valid but does not work for me. Either you get enough distance to get more of the sound of the piano (which comes back as room sound) -which means a great sounding room- or you try to fake it. While not ideal, we regularly have cases where the latter approach gives better results than the former. You can' win them all ...
What I would like to challenge is the use of LD mics, or any condenser mic for that matter: for me the attack is never clean, always artificial (compared to the real thing). A good ribbon like the SF12 seems to solve a lot.
Unfortunately, a Blumlein approach does not work at all in our magical hall. We are waiting for a cardioid or omni SF1 !