I think there’s also usually exceptions to the bumper rules for climbing if needed. For instance G23 from 2019.
Oh, there were quite a few swerve bots in 2016. But they had to invest mechanical engineering efforts into suiting their swerves for the game challenge that year, rather than using a COTS solution straight out of the assembly guide. 16 is one such example, both in terms of creating their module guards/wedges and also adding a number of other wheels between their modules to prevent them from bottoming out.
I don’t think anyone is arguing that 2016 is impossible for swerve or that FIRST should make a game that’s impossible for swerve. But rather, games like 2016 shift the design requirements in a fashion that require more bespoke engineering in order to apply a COTS swerve solution that addresses the game challenges as well as a tank drive does, and thus shifts the value allocations that teams do towards their decisions regarding whether or not they should pursue swerve for that game challenge. For example, 1640 is a team long recognized for their use of swerve (dating back to 2011) and sharing of swerve resources prior to the COTS swerve proliferation, and even 1640 opted against using swerve in 2016.