I agree with pretty much everything you've said in this thread, but there's a bit of scope for confusion in this bit.
I'm sure you know this, but its worth pointing out anyway just for clarity - (analogue) filtering is needed on recording before sampling, to ensure no frequencies above half the sampling frequency are present in the sampled signal - any that are there will be aliased down into the lower band.
--hjkl
You're correct, of course (aliasing distortion). I think my brain farted there or something. Here once again, the headroom afforded by recording at something like 96kHz or 192kHz could let you use lower order low-pass filters before resampling at 44.1kHz for CD mastering. Typically, this was "oversampling" and it wasn't really meant to be played back at those frequencies (there's a question of how much usable harmonics are up there from any given instrument anyway plus you can't hear beyond 20kHz regardless). On the playback side, the reconstruction filter or "anti-imaging filter" (also a low-pass filter design) is used to smooth out the digital "stair-steps" and reproduce the analog waveforms. "Oversampling" here achieves a similar effect in terms of filtering requirements. The primary difference is you can choose a playback DAC/Filter design, but as a consumer, you have no choice in what recording ADC they used. As long as a competent design is used, I'm happy. I don't claim to be able to hear these micro sub-0.1dB variations typically found in DAC differences.
Similarly, greater bit-depth while recording gives you a lot more headroom in terms of hitting the rails so-to-speak on your input gain with louder than expected sounds. To get maximum resolution, you'd want input gain set so your loudest sound is just below the maximum your amplifier will allow, but if something goes above that, you get clipping distortion (basically a square wave is created at the affected frequencies, which is very bad for ears and speakers alike). Go too low, though and you're wasting potential dynamic range as the softest sounds could have been more fully represented and without some filtering, your average volume output will be lower as well. Frankly, it's a bit harder to gauge maximum levels in a live concert environment than say in a studio, but it makes life simpler either way (i.e. more room to be off in the best gain setting since 24-bit gives you 144dB dynamic range and if your goal is only 96dB, you have 48dB of range to be off before you might clip). Of course, that means little if the mastering engineer is just going to compress the living crap out of the signal anyway so that there's nowhere near 96dB of dynamic range anyway (probably significantly less than vinyl in many cases, which typically peaks out around 55-60dB at most).
Even so, IMO a lack of dynamic range doesn't necessarily mean the sound quality isn't "clean" or "distinct" sounding. That is to say for example, while Tori Amos' "Boys For Pele" album has very good dynamic range and something like her "Scarlet's Walk" album probably has considerably less dynamic range, that doesn't mean that I think the Scarlet's Walk album "sounds bad". It sound clean and distinct and her voice is still natural sounding on playback. Her "Choirgirl Hotel" album, on the other hand is considerably more "grainy" and "rough" sounding to my ears. It's pretty squashed, but I think there's some other bad effects in the processing/mastering going on that make it sound flat and her voice less believable, etc. The point is that not everything in sound reproduction comes down to absolute frequency range and dynamic range. There's a lot more complex things going on along the way with modern mixing methods and DSP processing effects, etc.
Using SACD or whatever format this new PonoPlayer is going to use won't and can't negate problems present in the original recording, depending on the level of master available to go back and remaster. For example, I use Logic Pro and I have every input I ever recorded available with zero processing. But to fix something like processing effects that ruined the clarity of a vocal track or whatever, I have to go back and change or remove that processing. I can't just remix the levels of that track and expect those artifacts to magically improve. If I screwed up the input level even so and have clipping in the original recording, there's probably little I can do to correct it short of recording a new part. If it's just a tiny blip and non-crucial (e.g. during an instrument track), it could perhaps be edited in a waveform editor or even cut-out and then covered up with something more pleasant than a loud crackle, but your hands become more tied with what you can achieve.
Older recordings done on analog masters might even have more than one part going on per track in order to maximize the recording with a mere 4-track or 8-track setup. It can be difficult to address one part without affecting the other if they are concurrent. Processing effects may be hard recorded on the track (i.e. if you think using a slightly more pleasing distortion effect for a guitar part might sound worlds better, too bad, you're probably stuck with it if that's recorded that way on the analog track using an external pedal or whatever whereas in Logic Pro, by using Logic's digital "pedal" effects, I have a clean guitar signal recorded by itself on a track that I can go back and change to something else later on if I so desire).
The point is that even remastering old recordings might only get you so far. You're probably not going to be able to make a 1930's big band recording sound like it was recorded yesterday with modern studio equipment no matter how good of shape the available recording is in. In addition to the question of the limits of the recording medium used and the noise and other distortion levels present there, you also can't change the mics (both number and placement along with type) used or the ambient environment it was recorded in. On the other hand, I don't know what level of tools might be available. I'm pretty amazed at what can be done with a lot of patience restoring old film into a modern digital master. I'm not under the impression that sound editing is in quite the same league in that regard, though.
Either way, I'm still of the opinion that anything more than perhaps 18-bit on the playback side is overkill and frankly 16-bit is more than adequate for all but the most dynamic recordings (at least I'd never want to listen to music on a regular basis that had to be played at ear-damaging levels to hear the full limit of human hearing's capability for dynamic range. Anything but brief peaks in the 105+ range are hearing destroying levels, IMO and I typically try to keep my average playback level below 90dB for that reason. I do record my own music at 24/96, though for headroom reasons only. I haven't not been able to discern any audible difference between a final 24/96 mix output and the final 16/44.1 output (not hard to compare when you have Logic Pro running the original waveforms to compare against the final bounce.)