Staying with Alkan, the Etude Op.35 No.10, after ten or so placid minutes in G-flat major, suddenly concludes with a sharp, spiky couple of minutes in F-sharp minor using a slight variant on the G-flat major theme. Meanwhile, the Etude Op.39 No.2, after a coda in D major in which the piece seems to be scrambling to build up the energy to go out in a blaze of glory, ends up surrendering to its inevitable demise with three muted D minor chords. And then the finale of the Grande sonate Op.33 gets louder and louder as it plods through the final bars, to a crash in first C-sharp minor, then D-sharp major.... followed by three muted G-sharp minor chords, as though the anguish of the preceding bars is destined to carry on past the end of the piece (Alkan's intention, of course). It's also the only movement of the Grande sonate to end in minor - the first moves from B minor to the parallel major, the second moves from D-sharp minor to the relative major, and the third is in G major.
The three "Perpetual Motion" pieces by Poulenc all have peculiar conclusions as well; each seems to collapse in muted dissonance for the final chords.
The first movement of Beethoven's Sonata Op.10 No.3 in D major cuts off very suddenly; convention would have dictated that it end with three D major chords rather than just two, or at least that there would be a rest between the two chords that are there. But no, just two chords, right next to each other.
And finally, my personal favourite bizarre conclusion: Dvorak's Humoresque Op.101 No.8 in B-flat minor, the last of the eight Humoresques. Cuts off very suddenly in the final bar in the middle of a phrase used several times earlier in the piece, with a sudden crescendo to ffz for the final beat, which begins with an unexpected G-natural to end in the Dorian mode rather than Aeolian. (Which brings to mind the surprise Mixolydian conclusion to Chopin's Prelude Op.28 No.23 in F major - he even accents the unexpected E-flat.)