Note that here we're not talking about a single mutation but 17 non-synonymous (they change the protein) mutations with 3-4 with known effects on the interactions of the resulting proteins
What are you talking about? Sequence is unstable due to size? Coronavirus has proofreading mechanism that corrects most mutations due to errors, the estimate for mutation speed before this variant emerged was 2 mutations per month so this variant has ~10 months of evolution in them but probably evolved much faster for some reason. This is very much a surprise
The error rate is in terms of the number of nucleotides addended to the polymerizing RNA during replication
the mutation rate is primarily a result of RNA replication, _not_ time. this means more replication events leads to more "errors", this also means longer sequence has a higher rate of error than a shorter sequence. at a length of ~29k bp the coronavirus sequence is very error prone, so error prone that this is approaching the upper limit of stability.
it seems you are mistaking DNA proofreading for an RNA proofreading mechanism, which happens to be negative selection upon nonfunctional mutant/variant. there is no recognized enzymatic [rdRNApol] error correction associated with the coronavirus.
"the RNA polymerases of RNA viruses are the kings of errors – these enzymes screw up as often as one time for every 1,000 – 100,000 nucleotides polymerized. This high rate of mutation comes from the lack of proofreading ability in RNA polymerases. These enzymes make mistakes, but they can’t correct them. Therefore the mutations remain in the newly synthesized RNA." [0]
none of these error prone generated mutations are surprising, they are to be expected espescially when the virus is given opportunity to replicate at a high cyle rate.
this is very much not a surprise, if you are aware of the properties of +ssRNA virus.