This is exactly where opinions differ right now.

To my ears, the change from 16 bit to 24 bit is more dramatic than for 44.1 kHz and high samplerates. But you're right in one aspect. In today's music, compression is so massively used that some productions could be represented in two bits only (you gotta agree that dynamic range is lacking in some songs nowadays).

But where opinions really differ, in the samplerate dilemma, is why it sounds better. If you look at the Nyquist theorem, it proves that you can faithfully reconstruct a waveform if it was sampled at least at twice the frequency of the waveform. But Rupert Neve tried to prove something else played a role, and partially made a point. Take a 10 kHz sine wave. A pure waveform. It's in the audible range.
Now take a 10 kHz square wave. Theory states that a perfect square wave only has odd harmonics. Which means that the first harmonic present is at 30 kHz. Outside human hearing range. Still, if you listen to the two waveforms, you can clearly tell which one is the sine and which one is the square. So, Neve concluded, we can sense frequencies above 20 kHz. Is it necessarily by hearing? Maybe it is by such high frequencies hitting our skin... Could be...
In my view, this could be possible, I don't deny it. But I also retort that it is impossible to make a perfect square wave. So in this test (that was conducted quite some time ago), there were lower harmonics involved also, which can contribute in people hearing the difference.

What most people agree about is the anti-aliasing filter problem. If you sample at 44.1 kHz, you need to filter out frequencies above 22.05 kHz. But you must avoid touching at frequencies up to 20 kHz. You need a very steep filter for this. In the analog domain, steep filters induce heavy phase rotations, which you'll hear as loss of definition and sometimes distortion.
If you take a higher samplerate, the Nyquist limit (half the sampling frequency) becomes higher also. For example, at 96 kHz samplerate, in theory you could leave all frequencies up to 48 kHz untouched without having to fear aliasing. In audio, it's always presumed we can hear up to 20 kHz. So instead of having a very steep filter, you can use a much more relaxed one, you have lots more margin. Relaxed filter : less phase problems. So better sound. This right now is the main difference between "normal" and high samplerates. In my opinion, and lot's of others think also. It is possible to make filters that are very steep, and have almost no phase problems, but they are hugely expensive. I think a test was already conducted at 48 kHz, with a standard filter used, and a specially engineered "near perfect" filter. And in that test, most people were able to say which was the signal with the good filter. But almost no one was able to differenciate it with a high samplerate test.

Of course, we don't know everything about human hearing yet, so it might be that other stuff plays a factor. But another main thing you gotta keep in mind is that lot's of those high resolution quality claims are marketing strategies mainly.

What I know for sure :
- Real 24 bit audio converters don't exist yet. The best converters right now only achieve about 21 bit performance.
- It has been proven by Audio Precision that high samplerates don't necessarily mean better quality. High samplerates mean fast switching, and this stresses the electronic components very much. Stressed components don't perform at their best, so errors will be made. Right now it's safe to say that the best 16 bit 44.1 kHz converters still outperform most 24/96 (and certainly 192 kHz) convertors you find in soundcards.

Still, wether they are open to discussion or not, you have valid points.

But back on topic : all those explanations still don't show in what part the soundcard plays a role in the actual sound synthesis of the VSTi (or algorithm of the VST effect), except allowing you to choose a higher resolution in your software host (some hosts cannot work at 24/96 without having a 24/96 card installed for example) or exporting the sound afterwards through the analog output of your audio interface...