Another one is the integration of hybrid systems. For instance Ripple + Bitcoin (bulk of small transactions would move out of the block chain, basically things that wouldn't need to be register and/or have extra anonymity requirements).

There are a couple ideas on alternative data structures to store the blockchain more efficiently. This would mitigate both bandwidth and storage requirements by potentially an order of magnitude. (Although not so much for miners or general users, but mostly nodes and users of the "standard client" which is still the most used one).

So far nothing I've said requires a hard fork.

Now, forking (some of these are far from definitive solutions, basically only fee requirement changes are a real solution from the following ones):

Enforcement of a minimal but non-0 fee in the very protocol instead of by miner consensus. The way to go IMO.

Higher hard limit for the block size (kicking the can, but this can be a very effective kick that would very likely allow for the previous things to happen over time)

Equivalently, reduce avg. time per block.

"Floating block size" - not a solution proper but a mitigation system to avoid translation volume instability. It would adjust in a similar way difficulty does.

I don't know what's going to happen medium or long term, but the best plausible and likely scenario in my opinion is 1) some fee enforcement scheme 2) the bulk of small and local transactions to happen offline - both solving the latency problem and minimising blockchain bloat by orders of magnitude.