But even right now in bitcoin you can initialize a node with -assumeutxo to avoid IBD costs. You either get the UTXO set from a node you already control (and its turtles all the way down), or if bitcoin supported UTXO set commitments you simply pick a committed UTXO set from N blocks back, for some sufficiently large N. All block headers are downloaded, but full blocks back to the Nth block commitment are validated.

While technically SPV, this ends up being exactly the same security model overall, since an attacker that can invent (say) a full year's worth of alternate history to fool you can already perform arbitrary double-spends on main net. In addition, although this is not required, for most of bitcoin's history the last year or two of mining represents almost all of the proof of work anyway, so someone who serve you a more-work forked off false chain with a false UTXO set commitment could have also rewritten the block chain since genesis.

Bitcoin's security model already assumes this is not possible. So with UTXO set commitments (which bitcoin will eventually get in some form), you only need to IBD a few months worth of blocks, or maybe a year or so if you are truly paranoid.