So, as long as bitcoin scales well, the fee income for miners will never be able to rise to replace block subsidy
Although you ask a legitimate question, your argument does not prove this conclusion.
Estimates will be easier to make once it becomes clear where transaction volume is going and how much volume is handled by the bitcoin blockchain itself vs. how much is being moved to something like the lightning network.
This conclusion is based on the
fact that the fee per block is decided by the orphan risk, and the loss from a orphaned block is decided by its subsidy. So as block subsidy goes down, the loss from a orphaned block also goes down, and the fee per block also goes down
If orphan risk rises exponentially above certain threshold, for example 2MB block to take 10 minutes to broadcast to majority of nodes, then maybe fee will rise quickly to get close to block subsidy
It's not a fact actually, not for now at least. The current fee level is heavily influenced by default fees (0.0001 or so). In the future, when fees are fully market-driven, yes, they can be decided by orphan risk. But we also need to consider several proposals that can and will make orphan risks lower, and, more importantly, independent ( O(1) ) from the blocksize (provided miner cooperation). Things like IBLT. So, the problem is even worse this way.