Yes, the DPBSD Truepianos v1.5.0 test
concluded with a poor verdict. Yesterday I tried the current v1.9.5 demo and it was nothing to me to bother with. (In old times for 5 Ys. it was the very first piano sound I played with a MIDI keyboard and this first encounter was a true revelation. Interestingly, the timbre I found even now piano-like over the whole range, but after playing it some seconds you get all the problems which Dewster analysis is showing: stretching is obvious, a very audible phenomenon even during real playing.) Sympathetic resonance was non-existent for me, no much have changed since DPBSD test. (You cannot wait too much from a 100MB sampled instrument, but it passed our topic's test at least!)
I tried the current demo of Sound Magic Imperial Grand3D too. It has many of a piano characteristic implemented, but with many very unnatural rudimentary implementation details (reverbs, pedal behaviour).
Pianoteq stands there with no serious concurrency. Playability is superb, but it's piano sound I would describe as correct, but far from being a beautiful piano sound - not nearly as good as with some good quality sampled pianos, Galaxy e.g. (Even VPiano sound was not convincing.)
There must be a difficult to measure but subjectively very important distinguishing sound quality mark.
I can now better understand the strategy of modeler SWs, why they are packaging their SW-Product into whole phyiscal instruments: with SW alone it is no much room left to earn money. Even Pianoteq can charge only 100-400$ for their SW, setting some upper price limit.
What I could see as desirable, would be updateable modules within the cabinet of DPs. You could change computer parts (CPU, Memory, Motherboard, Sound Card all of them as a unit) AND SW separately, because the other parts (action, amplifier, boxes) are not amortising that fast. You have already standard interfaces to build on (MIDI; audio, sound cards).
I am wondering, how many LINUX-Based solutions are working within the DPs currently on the market.