I'd ground what I say in sharing your assessment of the observations or insight, but with a different takeaway.
If I need to think about the solution that hard, I may just as well take it all the way, remove the middleman and get some efficiency back.
Totally. And the takeaway is that we by and large should be doing this much more. That's what's needed from leaders. More vision, less delegation of vision. The stronger the vision, the less need for delegation anyway. There's something to be said about the difference between big picture and small picture, but at this point it's relatively intuitive that the big picture divorced from the small makes for worse big pictures. I'm not talking about how handing the big picture to the middle person tends to create interpretation/translation errors as well as execution errors to the extent the subordinate is less capable. Those factors are prominent and _in addition_ to the issue that the big picture is a worse big picture when these duties get stretched across two (or more) people. And thus, we should own these duties "vertically" and if needed reduce duties "horizontally".
Leaders (and companies) should be doing less horizontally. It runs the normal risk of stretching to thin, but their tenth and eleventh ideas are also filtered towards worse. Perhaps more fundamentally, the bar for adding a product/project should be higher. When the leader can "just" add another leader to take on the work—particularly one by nature with less power, ability, autonomy, and perhaps accountability—it's getting set up for a lower chance of success. In order: (1) the project probably shouldn't start at all; (2) if "started" at all, it shouldn't really look like starting, it should be a person validating for the leader who will then start it (many product teams pretend this is what they're doing); (3) if started, it makes quite a bit more sense for the higher (and likely more capable, but especially more empowered) leader to take on that project and hand the existing, stable, better trodded project where the team has institutional knowledge and support to the other leader.
In practice, sharing ownership is superficial, whether between leaders at different levels or between project team members. Why share? Own it or have someone else own it. Or if there isn't trust in the other person, cut it. If the leader doesn't have the time/interest and doesn't think they have someone capable of doing it, it should be as much a sign as a low quality idea.
It's not hard to find problems with middle managers—many are with the sorting that picks them, many are with the conditions of the middle. Structures and strategies to remove them, as you suggest, are a great idea. The main way is for the leaders, when faced with the scenario of "if I need to think hard", to not come away with "let's do it, but not me." We should be default no. Green lighting some work to validate it to a Yes works, but these are tasks such that the leader knows what kind of validation is needed—and is assigned to a person with those skills. An engineer, marketer, business analyst, user researcher, data scientist is going to validate things that a PM or a director have less or no training in. Leaders tend to appoint the latter though, sometimes for empire-building reasons, but I think more basically out of fear of control and legibility. Those are counter productive reasons if understandable.