- Full custody (Coinbase)
- 2-of-3 arbitration / DLC
- Threshold multisig (ecash, Liquid)
- Off-chain peg-out tx (statechains)
- Collateralized custody
- Lightning Network
- Pools/factories
- Aggregated coinswaps (Ark)
- Sidechains paper
- Drivechains
- Fraud proofs
- Softchains
- Availability sampling
- op_return / on-chain (Omni)
- Off-chain validation (RGB)
- Vanilla spacechains
- Fee-burning "two-way" peg
- Full on-chain data (zk-sync)
- Bit field data only (zendoo)
- Single device (opendime)
- Remote attestation (teechan)
- Probabilistic payments
- Plasma Cash
- Merged-mining
- Blind merged-mining
- Stakechains
- Proof of publication
This is a surprisingly uncommon opinion, but I stand by my assessment that they do not fit a useful definition of L2 (i.e. one that excludes extension blocks). Everyone on the L1 has to fully download and validate them and they cannot function an optimally efficient L1 that also uses ZKPs for validation.