Sure, fashion played a role, but in this case the main driver was the need to support devices of different resolutions. Scalability implies vectors, which implies shapes, fills, and strokes rather than bitmap assets.
As a designer who made "skeuomorphic" interfaces before the shift, supporting even a couple of different device sizes with hundreds of different PNG slices had already gotten unsustainable by 2010, and moving to a vector-based workflow was a breath of fresh air.
The pendulum did swing a little too far: some of the early human interface guidelines were wary of the concept of lighting (gradients and shadows). But, realistically, that was unworkable, and those guidelines changed because nobody followed them.
"Flat design" as a moment was actually pretty short. Later iterations on the concepts explored during that moment, such as Material, were actually thoughtful and workable, in my opinion.