I've been always a fan of committing to the chain by wallets for several reasons, as of your idea:
- There is no need for explicitly specifying the block height (your so-called flash-back-block field) as nodes could easily spot the target block, it is a waste of space.
- Speaking of space, It suffices to include like 8 last bytes of the block hash. To keep odds even higher, we could restrict inclusion of txns that refer to a block like more than one month earlier.
- I think there should be incentives for such a commitment and it suggests hashcash style work being included and accountend in a replace-fee-by-work fashion.
- Establishing an incentive mechanism helps to keep everything more smooth and implement it in a softer manner by making it an optional, new transaction format.
-An alternative approach would be committing to UTXO Merkle root instead of block hash.
There are a lot more to say ... will come back to this topic after a while at least for including links and references.