One client. One database. One Friday. That's it. If your migration plan needs more than a single screen of A4, you don't have a plan — you have a wishlist.
The only plan that survives contact with reality: freeze the legacy db, dump it, restore on the new version, fix what breaks. Everything else is procrastination dressed up as risk management. Schedules with three rounds of UAT and a "go/no-go" decision at every stage are just three rounds of paying consultants to wait.
The friction is the product. If migration is hard for you, your live data tells you exactly which assumptions in your business model are buried in the schema. That's the gold. Don't outsource discovering it.