The probable cause for many variants of concern has been multi-month long infections in one immunocompromised person (obviously a different one each time), often treated with convalescent plasma or monoclonal antibodies to keep the disease in check but not cure it. This provided the virus a perfect optimization platform, and for it to quickly collect a group of mutually beneficial changes.

In the normal case the selective pressure isn't really there, since the virus will get beat back after a week or two anyway and just 1-2 point mutations probably didn't give it any significant advantage is infecting more hosts.

Given how many mutations this variant has compared to its most direct known ancestor, the former is almost certainly what happened in this case too.

See for example: https://www.scientificamerican.com/article/covid-variants-ma...