On day one, Convex was architected for (among other things) eventual copy-on-write cheap snapshotting. We assumed we'd have built it by now due to customer demand. But we haven't. In practice, with our real customers, it hasn't actually ever made sense for them. Why? 1. If you have a large dataset you want to clone, that's typically "prod." In real companies with stakes, you can't dev on a copy of prod for PII/PHI compliance reasons. 2. If you have a small dataset, you're better off with…