So has the consensus on how Bitcoin is updated changed over the years?
yes
early on. new blank opcodes(no rules) were initially disabled, whereby rules were then applied to opcodes and then when there was somewhat majority readiness, they would activate the opcode to use such opcode AND enforce verification of a newly used opcode rule by having nodes ready first to know what is being validated at activation time(in short not allow usage of an opcode until network was ready)
some attempts of doing things "soft" resulted in forks which is where they then raised the majority from 55% to 75% to 90% and disabled opcodes to only activate at majorities.. (2012-13)
then they done the 2017 mandatory miner activation of august 1st with the block rejecting threat to push an activation, but only requesting miners and economic nodes to show readiness.. not user node majority..
they did however delay the wallet utility(tx creation using opcode) until 2018
the activation also allowed the creation of another layer of set of opcodes beneath the first set. some were disabled some were enabled(op_nop). where user nodes would just pass the active opcodes as acceptable without knowing the rules
and since then (2021) things have gone super soft where new opcode sets are not disabled upfront(op_success) meaning it doesnt require majority node readiness to enforce new rules and instead just pass whatever data is put after such opcodes unchecked. and its then for nodes to then code in a rule adhoc later, if they should choose to
strange thing is. many dont want to code in rules into the active opcodes, thus this avoidance of setting rules for opcodes means that it then allows any crap data to be put into blockchain data unverified/unruled, due to lack of rules in the active opcodes that should have been disabled by default
where they should only have become active once rules/formats are set for the opcode utility and only when there were nodes ready with the new ruleset to know what to look for.