A service can hit every target in its agreement while the underlying estate degrades. Response and resolution times describe how quickly a symptom was handled; they say nothing about whether the same symptom recurs weekly, whether the fix was a documented workaround applied for the ninetieth time, or whether the number of things capable of breaking is growing.
The second issue is knowledge. Transition is treated as a phase that ends, so operational understanding accumulates inside the provider and is not written down in a form the client could use. Over a few years this becomes the real lock-in — not the contract, but the fact that nobody else, including the client, knows how the estate actually works. Exit then looks risky regardless of satisfaction.
The third is that run and change are separated organisationally, so the team that operates the system has no route to fix the design that generates the incidents. They log the pattern, the pattern goes to a change backlog, the change backlog is prioritised on business features, and the incident continues indefinitely.
