[thih9]

We’d still need code to load the data back, also to continously test that code (or debug in production), plus handle data migrations. There might be more use cases. Using a regular db might be simpler at this point, depending on the project’s setup.

[curtisblaine]

Loading the data back is just reading the log file and applying every record one after the other (ignoring records outside the time horizon). Sure you will need to maintain the code, but it's not as complex as writing a database from scratch, as you suggested.