It isn't the 20MB block size that scares me, it is the automatic doubling every two years until we reach 20GB blocks. This scares the crap out of me because if a problem turns up we need a fork to pause the scaling.
I'm starting to wonder if forks are really that big a deal.
Imagine if all the exchanges were set up to handle economic arbitrage of two or more forks. Say you have 20 BTC on Bitstamp when a fork happens. Your account is automatically split into 20 BTC in BitcoinOld and 20BTC in BitcoinNew. If you don't know/care about the fork, you do nothing. If you think one of the two is obviously more viable and/or obviously more likely to get support, you immediately start selling BTC in one for BTC in the other.
Since everyone is doing the same, I bet this all plays out in a matter of minutes because once the trend becomes clear it will snowball since no one wants to be on the losing fork. Luckily, again, you can sit out the arbitrage and leave your stash untouched whichever fork wins. It's just that you can earn yourself some extra coin if you guess the winning fork correctly.
Now to guard against possible glitches in BitcoinNew, even if it wins in initial trading, BitcoinOld will probably still retain some value for a time - for instance 10% of its former value for a few days or weeks - as a representation of an estimated 10% probability of a glitch in BitcoinNew. After that it would likely fade into nothingness. All the while your bitcoins are safe no matter the outcome.
Not only is this far faster than waiting for "consensus," it also ties more solidly into the basic economics of Bitcoin itself. As Daniel Krawisz has pointed out,
where investors go, everyone else follows. Investors have the ultimate control, so the forking process should reflect this and exchanges should be setting their systems up for this.