The API of a sponsor-like mechanism is close to ideal in my opinion:
- compatible with non malleable transactions
- 0 overhead if fees accurately estimated
- watchtower friendly
- post hoc, requires minimal 'protocol awareness'
- friendly with most mempool eviction policies, not much new required
- can work to atomically bump multiple txns
- can be bumped cooperatively by multiple sponsors w/o coordination
- 0 'rebroadcast overhead' (e.g., for a large batch) leasing to cascading retransmission fees for replacement
- can be piggy backed with other future transactions or protocols (e.g. coinjoin)
- compatible with change being in cold storage
The main drawback is it is chain space - wise less efficient, as an additional transaction gets made. However, I think the API benefits 'product market fit' over alternative solutions outweigh other concerns, and if the 'sponsorship efficiency hypothesis' holds true, then most transactions will not require sponsors and therefore the savings of not needing to preplan a few bumping mechanism will be more efficient overall (efficient market will drive accuracy in estimating fees rather than needing to sponsor).
_______________________________________________