If our Liquibase changelog is corrupted (which is really easy to do when you're working with multiple workspaces, or if you're iterating on a table), deleting the DATABASECHANGELOG and DATABASECHANGELOCK doesn't work because Liquibase will complain about columns already existing when you add a column. (See https://github.com/Sage-Bionetworks/BridgeServer2/blob/develop/src/main/resources/db/changelog/changelog.sql#L207)
We need our Add Column changesets to be idempotent, so that we can easily recover from a corrupted database.
Side note: Deleting all database tables and asking Liquibase to recreate everything is a huge pain. This is because we have no Templates, so we'd also have to nuke our Studies and StudyConsents and flush redis.