The persistent backend storage adapter
Retrieve an entry from L1 (fast) with L2 fallback. Promotes L2 hits into L1.
Store an entry in both L1 (memory) and L2 (persistent) in parallel.
Remove an entry from both tiers in parallel.
List all keys from L2 (source of truth for key enumeration).
Clear both tiers in parallel.
Two-tier cache: fast in-memory L1 backed by a persistent L2 adapter. Reads promote entries from L2 to L1. Writes go to both tiers in parallel.