Production doesn't break at the moment bad decisions are made.

It breaks later — when enough good decisions have accumulated, made without systemic context.

Every contract optimized on price looks reasonable in isolation.
Every reduction justified by efficiency passes financial review.
Every replacement of a senior executor with a cheaper option seems like an acceptable trade-off.

Locally — everything adds up.

But systems aren't evaluated locally.
They're evaluated under load.

And under load, it turns out that a set of locally reasonable decisions created a system that can't hold pace, can't recover from failures, and can't scale without constant manual stabilization.

This is the classic local optimization trap.

In production systems it's especially dangerous because the gap between decision and consequence is often six months and an entire production cycle.

The evaluation metric must be systemic, not contractual.

The question isn't 'how much does this vendor cost'.
The question is 'what happens to the system when this vendor operates inside it under pressure'.

These are different questions. With different answers.