Implement all the features required for KV storage in Diskless log mode so that it can be used in production. In this case production means the system is stable enough to operate in diskless mode with no write ahead log on disk. This means that if more than a quorum of nodes crashes and the latest state was not snapshotted, data can be lost.

Some specific optimizations such as backpressure and log garbage collection based on memory pressure, and explicitly not included in this milestone. In order to keep this milestone tractable, log GC will only occur based on the global_min_accept value detailed in RFC 2.

Reconfiguration with warm nodes will also likely not be included, although this is up for debate.