The hardest decisions weren't visual. They were about whose pattern wins. When two products solve the same problem differently, one team's work becomes the standard and another team's becomes migration backlog. Handle that carelessly and you breed the resentment that kills design systems faster than any technical flaw does.

So we made losing cheap. Every convergence decision shipped with a documented rationale, a migration path and tokens that did most of the mechanical work. We picked battles in order of guest and operator impact, not in order of which inconsistency annoyed designers most. And we let teams keep local patterns where genuine product difference justified them, which made the mandatory parts far easier to enforce.

The forcing function was always the person at the end. When two patterns argued to a draw, we asked which one a tired operator at the end of a shift would understand faster, and that question settled more disputes than any style principle did.