I probably tried to make too many points in one post, and diluted my message by doing so.
My main thought was that moving from orchestra/string quartet/organ to piano brings in a whole host of challenges, related to the piano's inability to sustain at constant volume. (There's a whole bookful of tricks on how to do it - the simplest are things like replacing held notes with octave-tremolos.) A bundle of inverse problems arise when you arrange piano music for orchestra.
Accordions have a lot more in common with woodwinds and with organs than they do with pianos in how they produce their sound. On general principle it seems to me that orchestra->piano->accordion is just about guaranteed to both a) make the arranger's job a lot harder than it has to be and b) result in a final product that is less faithful to the original than it could have been.
Full disclosure: I consider myself a composer first, violinist (distant) second, accordionist third. I don't advertise myself as an arranger - I have done a little of it, by doing it well with someone else's music is hard.)
FFingers made some interesting comments, including this one:
"Also since many composers did/do create different versions of each work for various instruments/groupings/orchesras etc. just what is it that demands total dedication to those scoresheets which are presently available?"
In the 20th- and 21st-century context, the presumption is that the groupings on offer are the only versions you can legally perform without special permission. If you write and ask for permission, you may well get told no, or the composer may make a new arrangement for you, or may designate an arranger he'll allow you to hire; he is quite unlikely to let you roll your own.
150 years ago, the landscape was a bit different, in two ways. One, as there were no recordings, there was a big market for home sheet music, and writing a version of an orchestra piece for piano or small ensemble was almost like offering CDs for sale at a concert. Two, before international copyright treaties, Saint-Saens couldn't stop Liszt from performing his own arrangements outside France; his choices were to gently encourage Liszt to do something they both liked, or to hold his nose and let Liszt do whatever he pleased. (He was still alive when Liszt made his arrangement, and they knew each other, but I have no information whether they discussed Liszt's arrangement or not.)
Now, there are transcriptions -- attempts to be as faithful as possible to an original -- and arrangements, where the arranger is deliberately making certain changes. And either a third category, or an extreme version of arrangements, where someone takes old themes and makes a new piece out of them: here we have, for instance, Mozart's Don Giovanni, Hummel's Potpourri for Viola and Orchestra, and Liszt's Réminiscences de Don Juan.
My personal bias is that I either am as faithful as I can be (changes will only be to, say, simplify a piece for high school band), or I explicitly write a new piece. I don't have much patience for pieces that call themselves arrangements but add a bunch of original material, just as I don't have much patience for conductors (or soloists or theatre directors or movie producers) who claim to know better than the original authors .
I think it's abundantly clear that Saint-Saens's intention is preserved in the several variations he made of his own work, all of which handle things like bars 13-16 (D major chord held for 4 bars) the same way, and that Liszt wanted a piano showpiece not a literal transcription. (And our featured performer needed an accordion showpiece for the Coupe Mondiale, for a bunch of the same reasons Liszt wanted piano showpieces.)
