I really think that if bitcoin can't adapt to a very broad spectrum of events, like the mini crash of yesterday, another cryptocurrency will win the race. It's not that we have to win, by no mean, but we must continue to make bitcoin the best, with all we got witch is strongly related to winning the crypto-currency race.
So, the block size limit should be raised and should even adapt itself to the needs of the network. The only concern i can credit is if there is no limit, witch may brings centralization pretty fast.
The limits should be a function of the number of transaction pending in order that they get in the block chain with an expected waiting time of around 10 minutes.
i'm still new to bitcoin logic but something as simple as that could evaluate real time transaction size limit.
(Nb of transaction pending * mean transaction size (kb))/C
or if it is more usefull
(Size of all transaction pending)/C
Where C is a fine tune constant near 1 or exactly 1. if it is exactly one, someone will eventually confirm a block containing all the transaction (pending at time = some time) witch may not be what we want tough.
the idea is to accept a confirmed block only if, at time=t, the size limit chosen by the pool was under or equal to this hard limit. This way no miner can choose to work on a very large block and work out all the transaction by himself)