- Evict cache entries in order to provide bounded memory usage
- LRU cache?
- Hard limits
- Bound computational complexity
- Context-aware machinegun?
- Periodical history snapshotting?
- Reuse cache entries where possible
- Simplify cache interfaces
- Encode machine events and responses with a schema-aware protocol
- Thrift / compact protocol?
- Fix potential race when a
Repositoryrequest getting processed earlier than the start machine request being issued. - Don't go to the cache for getting
Headreference