The current disaggregated storage uses a precise checkpoint instead of a fuzzy checkpoint in the classic version to avoid performing a rollback to stable during recovery.

As part of this task, please check the current design that is implemented in the disaggregated storage prototype code and validate whether it is the correct design to perform a precise checkpoint and write down a doc to confirm whether it is correct or not. In case if there are any better alternatives, please list them with either pros and cons.