More accurately, it would weaken _trustless_ long-term value storage. People would be expected to trust the Armageddon block if the whole effort is supposed to bring any advantage. But many users use Bitcoin exactly because it minimizes trust in single parties.
The "Armageddon" block would essentially work like the checkpoints that were already part of Bitcoin Core since close to the beginning. Only that it would be nice to attach a (then-)current UTXO set to it, so the verification would be easier than if you had to reccur to archival nodes.
The "trust issue" for these blocks would lie in the question if they're really the longest chain tip at that moment, or if they are attack blocks (i.e. blocks that normally would have been orphaned). But that question would resolve quite fastly. If the Armageddon block has 1000 confirmations, the possibility for it to be reverted by a reorg is minuscule. A client version which would store a fake Armageddon block would lose all trust. (And of course clients should not drop all previous blocks once an Armageddon block has been recorded.)
Thus, it would perhaps not be fully trustless, but trust needed would be really minuscule. However, ZeroSync-style techniques to my knowledge would be much better because the proofs for the correctness of a certain block height would always be stored by all nodes, or all nodes using that technique.
The
mini-blockchain scheme (which is what the "Armageddon" technique basically describes), does work in some altcoins. However, it would perhaps create an additional incentive for destructive 51% attacks because the damage caused by such an attack would be higher, as you could try to rewrite history if the "common block history window" around the Armageddon block is too short.
