But simulation by bitfury already indicated that we will have a severe performance problem with 4MB blocks on average home computer
I'd like to see that report. Got a link?
With only several thousand running full nodes now, it seems to me that 'average home computer' is not the limiting issue.
I don't remember exactly, but just google it, and also another statistic by Mark in Montreal conference showing that even a 1MB block can take over 30s to verify on a mining node
Today we have the mining relay network, the block is relayed in milliseconds. What you are saying is factually incorrect.