The thing that could easily work if bitcoin were to fail as a currency is a currency which is backed by something -- computational value.

It's not very hard to securely run a virtual machine that could offer clock cycles from your CPU or GPU to some enterprise that needs them.  These clock cycles could be measured as VM process walltime and then be used to issue currency similar to bitcoin based on this.  Initially the currency should be extremely cheap from whoever is issuing it, spurring people to buy some of it.  The currency is then exchanged with users based on the wall-time of programs running on the VM on their computer.  The VMs will basically be like shells to the outsider who can run whatever they need too, eg parallel data processing.  The end user should be able to place restrictions on the shell (e.g. HTTP port access, mail server access) to prevent illegal activities such as spam.  In time the supply will be restricted and the people who want more of this currency to run more simulations will have to buy it back from the people they'd issued it to.

If we could figure out a way to string these nodes together for parallel jobs easily, this application could overtake super computers in terms of cost of utilization...

In fact, we could keep bitcoin going if we used it like this.  But anyway.

Bitcoin I think is set to head down to $4-5 each, I give a detailed explanation here: https://bitcointalk.org/index.php?topic=34978.0
Basically the era of 100% ROI is over for BTC