An end-to-end test detects issues whenever there is a major system change. In software development, with every release, a battery of unit tests must pass or business can be lost.
Yes, software testing is difficult. It sounds like it is reactive but it is preventative.
Do you have a defined end-to-end test path you can implement in a code replay – a commercial one is Selenium – you can make it record your “happy path” from home-page to shop order to basket acceptance to payment page, do you have certainty that the new plugin you are testing will not break something you rely on?
The reason for code isolation is to prevent this, but WordPress operates in a global namespace – know that the basics are not broken by a plugin update. If it is a server-side script, it can for instance test certain URLs work (return 200 OK instead of 404 or more to the point when it returns a 500 or a 502 HTTP error in a plugin function you may not know something is failing).
With a “headless-browser” on the server, we can automatically test it is working. It could simply be an app that tests URLs that you can visit, or paths you can get 200 OK responses for – rather than a WordPress plugin that has overhead on every request.