An important consequence of reconceptualising group selection in terms of complex systems theory is that group selection no longer requires competition between groups.
this was a serious constraint on old concepts of group selection, because it was a stringent criterion for groups to replicate, over multiple generations, the kind of competition which is so often seen between individuals - with additional problems of group cohesion, and the need for the spread of adaptive mutations (beneficial for inter-group competition) within a group.
Instead, group selection is assumed to occur whenever there is a sufficiently dense network of communication and interaction between individuals sustained across a sufficient number of generations (how many generations are required for group selection to have a significant effect depends on the density, hence selective strength, of the system of communications) - remembering that 'the system' is abstract and consists of communications, and the system is not the physical individuals.
(The physical individuals are termed 'communication units' and are strictly regarded as outside of the abstract system - not a part of the system - but their function is instead to generate and receive the communications.)
So group selection can occur even when there is only one group.
Group selection should therefore be conceptualised as something that happens individually, to each group - rather than as a consequence of competition (or any other form of interaction) between groups.
The divergence between group characteristics is a consequence of the different selection pressure operating on each group - as a consequence of each group having a different system of dense inter-communications.