Funny you mention this, I was just writing up some of my thoughts on this.
1.) Detail the Statement of Work
2.) Write the tests & Acceptance Test Procedure to satisfy the Statement of Work
3.) Run the Acceptance Test Procedure in completion from start to finish with no interruptions.
4.) Document the testing
5.) Any changes to the system require a full system pause until a Successful Acceptance Test can be validated.
This worked for a complex system flying large unmanned aircraft with payloads over civilian areas. What I see frequently is that a change is made, and only the area that is affected it tested and something else breaks since in many cases services have not been properly isolated and you end up with somewhat of a Ruth Goldberg machine. And this is the case of developing things in an API First approach that has a requirement of automatic testability. It's more work at first but once you get used to it, it's faster since you spend less time chasing bugs.