The spike protein also consists of two subunits, each of which consist of multiple domains. Some of which are even known to be able to fold autonomously, without being part of the entire spike protein(in particular the receptor binding domain alone can do this, and it’s just about 200 amino acids long). The RBD alone has about 15 mutations total.
Whichever way you look at it, this protein should be pretty much half-way if not entirely dead if Miller’s confused assertions reflected reality.
Now the real problem with his assertion is his misreadings of the literature he cites. Particularly what he takes the Tawfik study to imply.
It is true that the average effect of mutation without selection has a destabilizing effect on protein structure, and it also makes sense that the magnitude of the destabilizing effect of each individual mutation would diminish as the protein grows larger (because, basically, the larger the protein the more internal surface area can “stick” together so any individual amino acid would be a smaller part of the structure). But Brian Miller takes this to imply one protein can’t evolve to acquire a new function (in the video he basically asserts that the flagellin protein of the flagellum couldn’t evolve from a protein with a different function) by accumulating mutations, because he appears to not understand that purifying selection can maintain structural stability of a protein against destabilizing mutations, such that you can continue to accumulate mutations well beyond the naive threshold you would reach without selection. His idea seems to be that if two proteins are too different from each other in sequence, then there is no way to get from one functional protein to another even if they’re as similar as 90%, because by the time 10% of the sequence has changed the protein should be completely nonfunctional.
That is absurd.
We know of examples of proteins that diverge in sequence basically all the way from 100% to 15%(or even below in some cases) sequence similarity, while retaining a functional and stable structure. And we have divergent superfamilies where we know of variants that diverge at large portions of the way, at increments on the level of 1-2% at a time. 100%-98%-96%-94%(…)10%-8%-5%-3%. Incidentally there was a good example in table 2 of the paper I referenced in this thread(and again here).