Drives are cheap enough, and if you are running a node that is doing processing for yourself / business then you should have good enough bandwidth to handle a few meg burst here and there. With compression it's getting even better.
I have a client running a eth node AND a BTC & Lightning node on a 768/128 DSL line with no issues. Needs many TBs of SSD to run it because of the ETH but don't tell me you need massive bandwidth to do it after the initial sync, I can drive you out to his office in Islip and point to proof that you don't.
Drives are cheap, yes. Not everyone can upgrade or have space for upgrades. Laptops are fairly limited in disk size and it can be a hassle to upgrade to a bigger drive and there are plenty of heavy users who just needs that disk space. Bandwidth isn't the main problem. The time it takes to propagate and validate the blocks will be the main problem. Having a long propagation time introduces the problem of forks and they would have a lower security when getting any confirmations.
As to how much block size should the increment be, I'm not too sure. Segwit is a block size increase, it's not right to say that the block size hasn't been increased before. Though with the current mempool, perhaps 4MB blocks wouldn't be enough.
And yet here I sit running 2 of my own lightning nodes for my own use, so I can fill them when fees are low and slowly spend them. But I still have to do on chain transactions because a lot of places still don't take lightning payments.
Unfortunate. I find that a common issue as well.