Blockchain pruning is entirely do-able.

We've already had several altcoins launch (most recently Clams) with an enormous genesis block that contained a "snapshot" of current bitcoin txouts as of the time the genesis block was composed. 

Bitcoin could also switch to a new genesis block.  Here is how it would work.

1. New client is created and starts publishing a new block version, indicating a "hard fork" is coming.

2. New client waits for 95% majority on its block version, then starts publishing a new block format, which contains a hash of the current set of unspent txouts.  This TxOut set is "stripped" of contextual information including its block height and the transaction in which it originated, just as the "spin-off chains" among altcoins are made. Essentially it's just amount and key hash. 

3. Everybody can check that the new blocks are descended from the original genesis block (continuity hash) *and* that the entire set of unspent txouts is one that agrees with their current image of the universe.

4. One thousand blocks later (so well after any possibility of an orphaned chain), a special "MegaBlock" goes into the block chain.  It contains the set of unspent TxOuts that corresponds to the hash that was created for the first new-format block.  Thereafter, TxOuts created before that snapshot may not be used in transactions; however, a wallet containing the key that would have spent such a TxOut can now use the same key to spend a corresponding unspent TxOut that's published in the MegaBlock. 

5. After the MegaBlock everybody can still use the existing continuity hash to ensure that the block is descended from the first of the new-format blocks, and use the MegaBlock itself to verify unspent txOuts.  And it's up to them whether they keep the block chain back to the original Genesis block or not.

6. With the new block format containing a hash of the current unspent TxOut set, everybody who's following along can check that their TxOut set matches the hash that's published in every block.  Additionally, a new MegaBlock can be published every few years, allowing the block chain previous to it to be dropped.

This could be done differently:  Instead of having a Megablock the new block format could just reserve some space in every block to publish replacements for the oldest unreplaced txOuts, enabling the blocks prior to the block containing the last replaced txOut to be dropped (a "rolling root").  That would be more efficient if the blocks containing the txOuts are uniformly more than 2 years old.  But, all at once and nothing first is a heck of a lot easier to check and be sure you've got it right.

Anyway, my point is that blockchain pruning is not a technological risk; it's something that there are known and implementable ways to do.  I've outlined one such way, and as protocol, it checks.  It's not terribly efficient, but heck, we've been broadcasting every transaction twice for years now so efficiency is probably not a showstopper.