True, also there is no incentive to actually operate such a node and if they get rarer, they actually have to serve even more data which puts a stronger pressure to shut them down.

In the OP however there is also some BitTorrent distribution mechanism mentioned... even though torrent files (actually info hashes) for a given file are not 100% deterministic, they can be made deterministic, if there is an agreement on block size. My suggestion would be to include the info hash of a bootstrap.dat file that contains all blocks from (#currentblock-120-#currentblock mod 2016) - meaning that120 blocks after every difficulty change a new info hash gets posted to the UTXO chain. That way the torrent file stays static enough to be properly seeded but changes quickly enough for new clients to catch up.
As BitTorrent downloads are properly hashed, checked etc. and the blockchain can also be verified again upon importing, this should be safe enough, if one also trusts this UTXO chain.
That way full archival nodes would not be queried for older blocks that much, as the bootstrap procedure for a "Ultimate compression" client would be:

* Get current block height
* Get block height of the highest UTXO block X blocks below current height (I would stay with 120 as this is the magic number for spending coinbase transactions, not over 1000... but this canbe even user configurable)
* Download this UTXO block and all UTXO blocks after it, as well as all mainline blocks since then
* Verify that following UTXO blocks were correctly created and also that mainline blocks are OK.
* Get all mainline block headers since genesis (probably a big art of this data can be already come with the installer) and verify that the recent mainline blocks are part of this chain.
### Here you can already start mining your own blocks ###
* Get the full block torrent info hash from the header (?) of the UTXO block(s)
* Download the torrent via a magnet link over DHT (all you need is the info hash after all)
* Verify all full blocks since genesis
### Now you are at the same stage as current bitcoin-qt ###

Now you can either continue to seed the torrent (and maybe even do this on a dedicated machine!) or be happy that the UTXO chain was actually not lying to you, discard most older block contents (maybe here it might make sense to keep the most recent 1000 or so in case there are longer reorgs somewhere in the future?) and just work on UTXO/mainline chains.

With BitTorrent it would be possible to help distributing the block chain without running bitcoind, a downside is that any client that does not have the info hash of the most current check point will seed a relatively dead torrent...