I'm a software engineer in a research environment. GP sticks out to me because it's nearly the opposite of how it works for me. I as the architect/lead software engineer make basically all of the important decisions about the software. I have a project manager and about 6 researchers who are, theoretically, my customers, but they all know that they're not software engineers so they tell me what they need and defer all intermediate decisions to me.

Combining my experience with GP, the glib, almost tautological, answer is that you fix the problem by not letting "non-technical middle managers" make technical decisions. Actually implementing this in an organization that already suffers from it is a political problem which I don't have any advice on. For orgs where this isn't a problem yet, the answer is simple. If you're a technical organization, don't hire non-technical managers. Every single person in my management structure started out as an engineer or researcher and it works remarkably well.

Enterprise architecture done well; but the tooling is incredibly difficult, it's expensive and you need a culture devoted to engineering discipline over raw productivity. Other industries manage this just fine: manufacturing and defense contracting are two that come to mind. Both industries have pretty good work / life balance for their employees.

Tech is just worse because we haven't unionized like both of those industries did, so the companies are able to operate in a way that is more exploitative of its workers. I think that's actually the source of the outsized margins we see in tech. Agile is just shorthand for "it's too expensive to coordinate things so we'll just make our engineers do that on top of being engineers". The solution is to unionize and start to demand some standards in engineering discipline that allow people to have both a career and a life.