In theory, RW can support bootstrapping historical state if the data comes from another RW job, but it requires significant design and implementation effort.
If the historical data comes from a raw table in the data lake, such as a Hudi table, encoding the data in a custom format and storing it in the storage engine for later use is necessary.
RW incrementally computes, so bootstrapping historical state may involve processing from scratch to derive a consistent internal state along with other intermediate states.
