I'm not sure why you guys are now speculating of a future in which a 100k increase of the block size limit is possible. You are not factoring in computational limits into this. Even if storage and bandwidth were very cheap (affordable for such blocks) and everyone had connections that could support them, there are still computational limits. Validating such a chain, or syncing up to the network would be a nightmare (you would likely never be able to catch up).
And we would need such speeds across continents, dedicated entirely for the support of Bitcoin infrastructure
The problem with increasing the block size is not only the transfer speed.
Yes, Segwit will effectively allow larger blocks, but that's not its main purpose, and the increase is marginal.
The most recent transaction type data shows that we would have a probable block size of 2.1 MB (up from previous 1.7 MB calculation). So 2.1x is what we call marginal today?