This pretty much sums it up. Narrowing the scope of maintenance is a big deal. Once I'm confident in the quality of a framework, the next question is how likely

This pretty much sums it up. Narrowing the scope of maintenance is a big deal. Once I'm confident in the quality of a framework, the next question is how likely and how often am I going to need to use escape hatches for this project now and in the medium term. If I expect too many (or too difficult) escape hatches, it's the wrong tool. If I can't find a framework that I'm reasonably confident in that way, only then would I go without. If so, whatever I come up with is by default a framework. It's just probably not going to be as good as an open source one.