This generally happens when a virus from another animal (and thus with a slightly different version of the receptor) adapts to humans.
Basically, a gradient descent from a very different starting point ending up in a different local minima.
The question is, is the spike protein at a local minima (very probably quite close to it), and could it jump to another local minima then? Most likely not.
It's possible it will evolve to make our antibodies slightly less effective, likely as a trade off for infectiousness, but incredibly unlikely the spike protein would evolve to be unrecognizable to our immune systems. Historically, this doesn't happen all that often, except for viruses with chronic infections over years that can do a deeper search over the gradient, but even then it generally takes years to decades.
Generally though, within the context of machine learning, one of the benefits of gradient descent, especially when stochastic, is that we can get past those local minima humps. Does this hold less true with respect to the process of sequence mutation that viruses go through?
The ancestors of this virus were bouncing around between a relatively small set of humans and animals for a while before becoming SARS-CoV-2.
Now with a significant portion of the world infected, it does not seem unreasonable that it would mutate. Although at that point it probably would not be considered the same virus anymore.
This is effectively impossible as the virus is dependent on a functioning spike protein in order to infect cells, and the machinery involved with the mechanics of the spike are extremely delicate. Errant mutations that would cause compositional changes to the shape of the protein are almost guaranteed to cause functional failure. Using terminology, this is what we call a highly conserved area.
In both cases for the South African or UK strain, if you take a look at the areas where mutations have occurred, you’ll see that the code responsible for generating the spike protein is basically completely unaffected. This will generally hold true for any successive future strains.