A look at Arbitrum AnyTrust Chains

Layer 2 (L2) scaling is a general framework for alleviating computation or storage loads on a main chain (L1) to boost application performance. Most L2 solutions are operated by designated L2 nodes, which perform state execution computations in a more centralized, higher performance environment not suitable for L1. 

The tradeoff of centralizing state execution on L2 is considered acceptable if honest behavior is guaranteed by mechanisms on L1 (fault proofs for OR, validity proofs for ZKR). However, due to the elevated risk of downtime on L2, it is also necessary for L2 implementations to support transactions with guaranteed uptime via L1 execution (e.g., escape hatches).