Where are you getting your data that we have seem a 1000x increase in bandwidth and network propagation times between 95 and 2015?
I was actually using the internet back in '95 so my personal first hand experience is the data.
To give you an idea, back in '95, my entire country was connected to the Internet through 2 major ISPs. Their total connectivity added up to ....512 KBPS, which had to be shared to >5000 users who were using dialup (typically 14.4 to 28.8 kbps - although a bad phone line could drop that rate dramatically) or leased lines (64kbps). Today total connectivity is in line with the 1000x figure (at least 500 gbps). My home connection exceeds x1000 gains if I connect to a VDSL and is at 500x-1000x with ADSL.
The exception that breaks the rule.
Ok, well you were under a very specific circumstance of a country that didn't adopt the internet as quickly. Keep in mind that in other parts of the world things were much different.
Actually I was far better (in Europe) than most parts of the world.
The only place where speeds were better than Europe was the US. But that didn't involve international traffic (which is what the Internet is about), only domestic traffic - because, in a sense, the majority of Internet content, websites, etc were mostly US-based so the traffic was local to the US.
Still, back in '95, dialup modems were pretty regular for home use in the USA too. There was no ...FTTH and ADSL was on the ...horizon:
http://s29.postimg.org/q0hlsbb8n/DSC00023.jpg (Byte / Jan 1996)
so If we were to use just fixed bandwidth the numbers would be close to 50% per year... so would you be ok with a 50% a year blocksize increase? Of course , IMHO , we should be much more conservative than that because propagation time and network latency isn't also increasing at the same rates and really important networks like TOR are also not increasing at the same rates.
You have to remember that user adoption is on a curve which is much faster than technological progression, so less than what the technology is able to give you won't cut it. But allowing spam won't cut it either. So I'd like to have as much as possible, the sooner it becomes technologically feasible, with one slight twist: Serious minimum fees that prevent network abuse for the lolz.
As for latency, latency of miliseconds in a network that operates in 10m windows, is relatively ok. If we are talking about propagation delay, then transmission speed (which has the tendency to scale) plays a huge role. If you have networks that transfer files 10x as fast, you get the job done quicker, thus lowering the propagation time singificantly.