But with hard-fork frowned by majority of Bitcoin community, the choice at that time is either that or keep using bytes (which means no block size increase).
they argued "hard fork is a crime" when they said bitcoin needed to be qt0.1 compatible
funny thing is QT0.1 would fail as soon as it synced to 2013 data
as for modern bitcoin. easily having code that at block 870,000(example demo number) if block version is 10 then txid merkle is for all transaction in a 4mb space
where it only activates if 3630 of 4032(90%) blocks show version10, else do nothing
giving a year for people to upgrade node
and yes its possible core went from core 12 to 14 in a year back in the leadup to segwit and no one was using ant qt versions