The good thing is that, according to digital sampling theory, it is possible to rebuild the exact initial analog data from the sampled data by using an ideal (sinc) interpolation function. This means that you could indeed "extract more data" from a digital file (redbook or higher resolution) by filling the gaps between samples with zero error compared to the original analog signal which was recorded.
The bad thing is that using an ideal interpolation function would take an infinite computational processing capability, so we need to live with a certain error, and all DAC designers have their take and preferred approach to how to reduce this error as much as possible.
The ultimate achievement of the MS (+dual BNC DAC) is to rebuild the original analog signal with an accuracy better than 16bit (16.6bit actually).
This in turn produces a timing error which is close to - or better than - the threshold of human brain (a few microseconds according to literature) of discerning transients.
In order to do this, the approach selected by Rob is to use a very long (1Mtaps) sinc function interpolation filter, and to upscale the sampling rate to 705.6/768kHz.
The true "magic" of M Scaler resides on using its full 1Mtaps filter length / WTA combination algorithm, because only by doing so you would bring the timing accuracy to a level where the brain
should not be able to discern transient errors. RW insists that transient errors are at the root of all evils for audiophiles because they degrade also timbre and bass pitch perception.
Therefore, to answer your questions based on my (admittedly limited) grasp of all this matter:
a- yes, in a way (see above) the MS 'extracts more data' and 'lessens the data that's lost' from files with resolution lower than 705/768kHz;
b- if you had a native file of a certain resolution (say 24/192), and use the MS by starting from a lower res one (say 16/44) to upscale "only" to the native level (i.e. to 24/192 in this example), the MS would not produce any advantage, as I understand it
c- in today's world, native files at 705.6/768kHz are not easily available to us consumers, so by using the MS basically all practically available (PCM) digital music can be improved to some extent
As an M Scaler owner, I can testify that I can hear improvements from the HMS even when playing high res (e.g. 24/192) files, and that there is a marked difference from using the HMS at full throttle (1Mtaps, hence everything upscaled to 705.6/768kHz) compared to using its upscaling capabilities only partially (which you can do by adjusting the settings of the HMS OP SR button to passthrough - low - medium - high).
In addition to the videos already linked by
@Triode User, here you find some other interesting material:
https://www.head-fi.org/threads/watts-up.800264/page-12#post-13150760
https://chordelectronics.co.uk/wp-content/uploads/2018/07/The-theory-behind-M-Scaler-technology.pdf
Hope I have not posted anything misleading, if so other more knowledgeable people and of course Rob Watts can straighten this out.