It is a question of incentives. Subsidies incentivize honest mining.

As subsidies erode, the incentive for basic transactional block inclusion starts being eclipsed by exogenous incentives for shenanigans (a.k.a., MEV). Anything a miner can do to get sats (or fiat/altcoin bribes) starts to add up relative to the simple process of advancing the chain in an honest way (block subsidy).

As fees increase, there is more incentive to reorg the last block and/or start mining a second new block before broadcasting a new block. Why? Because side-channel bribes and greedy fee capture maximization start to overshadow the basic incentive to advance the chain (a.k.a., the block subsidy).

The reason you may see increasing discussion on this matter is that the whitepaper does not account for value accrual outside of the block reward, and it assumes all blockspace only exists to facilitate transactions. Inscriptions, OP_RETURN, mempool.space Accelerator, and other forms of side-channel value extraction seem to invalidate the security assumptions discussed in the whitepaper, IMO.

I think until new maths and new models are formulated to explicitly prove security under modern operational timechain conditions, the legitimate concerns about future finality will continue to increase.