You say UTXO commitments is "a strict reduction in security". If UTXO commitments were rolled in as a soft fork, I do not see any new attacks that could happen to a person trusting the committed UTXO + any remaining blocks to catch up to the head.
I would imagine the soft fork to proceed similar to the following.
1. Miners begin including UTXO commitments.
2. Miners begin rejecting blocks with invalid UTXO commitments.
3. Miners begin rejecting blocks with no UTXO commitments.
To start up a fresh client it would follow the following.
1. Sync headers.
2. Pick a committed UTXO that is deep enough to not get orphaned.
3. Sync blocks from commitment to head.
I would argue that a client following this methodology is strictly more secure than SPV, and I don't see any attacks that make it less secure than a full client. It is obviously still susceptible to a 51% attack, but so is the traditional block chain. I also do not see any sybil attacks that are strengthened by this change because it is not modifying the networking code.
I guess if the soft fork happened, then miners began to not include the UTXO commitment anymore, it would lower the overall network hash rate, but this would be self-harming to the miners so they have an incentive to not do it.
Please let me know if I have missed something.