People have discussed sidechains and pruning and other suggestions, but the fact is, they will take time to implement to make sure they actually work. A 20MB blocksize is a far more simple solution that will work right now and until I hear something better, that's the one I'll be supporting.
Ok, but, see, we can go for 2-5-8 MB limit (wihtout exponential growth) before we hit the limit, and use the time gained by shooting the can to invent something better
It's not so urgent that our only choice is to implement exponential size growth
What you are saying now is, let's go exponential, maybe we can sustain it.
Blocks won't be instantly full when the limit is raised, so there's no point in going with not so big blocks. Unless you have another argument for doing it.