The primary argument to retain the 1MB limit is that smaller blocks helps ensure decentralization of Bitcoin. Lots of people rightly care about this.

As the average block size has increased the number of full nodes has been steadily declining. Once decentralization is lost then it is unlikely to be recovered. So this is a one-way path when changes are made which put decentralization at risk. The number of full nodes is in flux as they keep coming and going. Maybe there are about 8,000 at any one time. Ideally, this should not drop too low.  So, there is a trade-off between average block size and the number of full nodes. No one knows the whether there is a tipping point where a certain block size might suddenly cause a lot of nodes to disappear because of the volume of traffic. Consumer bandwidth is the key limiting factor.

The best way to keep blocks small is via fees, and the minimum fee was increased 1 year ago which got rid of a lot of the messaging spam which had tx volumes at >60k per day early in 2013. So fee increases have already allowed the block limit to remain longer than previously expected. The block limit has a use for its psychological effect of driving the fees market, competition for block space. The problem there is that the fees market may not work smoothly until the block reward is 6.25 or even 3.125 btc.

However, a fixed block limit is very dangerous because of the potential for ecosystem conditions to create a sudden volume spike into this bottleneck crippling transaction throughput, collapse the price and create a PR disaster which will hang like a cloud for years afterwards.

IMHO, at the very least the 1MB should increase in line with the global improvement in bandwidth speed, and should be about 3MB by now.
This can be done using block version numbers and a super-majority trigger, or the more primitive block height check. However, time is running out and it looks like the 1MB will be maxed out during Q2 2015.