To do this you would basically need to just remove the liquibase tables from the database and then create a new changelog. Personally, I’ve found that the benefits I get from this type of ‘compaction’ are not sufficient to outweigh the costs of doing it, but I’ve mainly done this on a fairly small project, where we only had ~300 change sets. I could see that if you had several thousand and the time to create a new clean database was getting very long, then it would be worthwhile. I think about it in similar terms to how I think of build times. If the build time gets above 5 minutes or so, it is worth spending the time to optimize the build, but decreasing build time from 1 minute to 30 seconds usually takes more effort than it is worth.