Yeah, it is subjective. My desktop could easily handle 20 mb blocks, and it isn't anywhere near top of the line.
No, I'm fairly certain that you've pulled that out of thin air. Exactly where, when and how did you benchmark validation time for a 20 MB block? Quadratic scaling induces problems with malicious 2 MB blocks, now imagine the implications on 10x of that.
Sure, in 5 years I'd need more disk space, but I could also buy enough disk for 20 years at that rate for less than $600. In 5 years when I actually need it, that storage will likely be even cheaper. The CPU, RAM, and network connectivity I have today can already handle it.
You are forgetting several things in your point of view: 1) Syncing from scratch; 2) Bandwidth. Sure, you could argue that you'd need only download 20 MB worth of data on average every 10 minutes. However, this does not apply for full nodes, but full clients. A node, doing an average of 1 mbit/s at the current block size of 1 MB, will spend around 300 GB of data in 30 days (source: My own node).