The top three reasons are perfect examples of worst possible assumptions, on which no sound engineering (especially of decentralized consensus-critical payment network) can be based.
Once again, BIP 101 is not the end-all be-all. It's a first stab at the problem, and a reasonable one at that, imo.
But let's take the worst case scenario: technology ceases to improve whatsoever, transaction volume skyrockets enough to fill all blocks 100%, and all the miners (irrationally) choose to jam every block completely full. In that world, a reasonable desktop PC running on a home connection, and with hard drives that today cost about $300 will have zero problems keeping up and storing the data, and operating as full nodes, until at least the year 2023. In that made-up, dystopian apocalyptic future, we would have to (gasp!) consider a fork... eight years from now.
Is it really reasonable to assume zero technology improvements in the next eight years? Of course not.
(Table 2). Currently, in the worst case scenario, my PC
handle 8Mb blocks, but I wouldn't bother running a node that consumes 1MB/s of traffic and eats more than 1GB space a day.