Your modernisation isn't a rewrite. It's forty small evictions.

Every big-bang replatform we have been asked to rescue failed the same way. Not on the technology — on the calendar. Eighteen months in, the new system did most of what the old one did, the old one had kept changing, and nobody could say which was authoritative.

The strangler pattern wins because it makes that question impossible to ask. There is one system. Pieces of it quietly change hands. A request that used to hit the legacy path hits the new one, behind a flag, for one percent of traffic, and the rollback is a toggle rather than a weekend.

The sequencing matters more than the technique. Start with the piece that is cheapest to move and most expensive to get wrong — usually a read path, rarely a write path. Reads let you run both systems in parallel and compare outputs without consequences. Writes force a cutover on day one.

Instrument before you touch anything. On the uNation engagement, two weeks of tracing showed the real cost sat in a fan-out query nobody suspected, which rewrote the plan we had been hired to execute. That is not unusual; it is the normal outcome of measuring first.

Budget for forty evictions, not four. Each one is small enough to reverse and boring enough to do on a Tuesday. That is the whole point.