when you learn how cores fee estimator works, then you will see how your notions fail

core doesnt count fees once tx confirmed.. it instead uses the UNCONFIRMED mempool transaction fees and how delayed they are of even getting in a block to estimate the fee new transactions should try to push a tx forward

meaning if the fee was high and pools declined to accept transactions below a certain rate, even if it meant doing part filled blocks/empty blocks... users wont be happy.. the more delayed a tx gets the higher fees go.snow ball effect because users wont transact if costs are high and still not getting in. and if pools are demanding high fee it helps no one

your "if onchain transactions go down, this favours users" doesnt play out if users are still paying huge fees which still sitting in mempool

as for the debunked comparison idiots make about visa.. please do get a new song sheet
only idiots use the visa song as a reason to suggest bitcoin shouldnt scale at all unless its to those extremes.. which isnt even scaling.. its exaggeratingly leaping.. and debunks itself to use leaping song to avoid scaling discussion/options

and you then go extremes of the opposite about not scaling at all, to then say it wont be secure
..and your other illogical fallacies...
you have been trained very (un)well to only assume the two options are do nothing or exaggerate to a huge leaping massive size

..
anyway back to my point
if the only users that can afford huge fee's enough to get accepted into a block are the scammers that overcharge victims on a scam market that 1sat is worth $10k, and need to pay $100 in fee to get the special sat.. then normal bitcoin users will stop transacting.. and all thats left is the scammy crap
(one scammer was willing to pay 83btc in fee just to get coins he hacked moved quickly)
no one will be able to outbid scammers.. and we should not condone the crap that end up penalising/costing all genuine users utility