Most IRL software failures like poor performance, lack of stability, unmanageable technical debt, etc are going to be systems and human failures 90%+ of the time, not the individual tool or framework choices.
It’s human failure 100%. Big frameworks just lower the baseline such that this failure becomes more acceptable, less immediately aware, and masks candidate selection.
If you want to achieve something smaller or faster you have to measure for it. Odds are if you cannot execute without a big framework telling you how to proceed you will not be able to measure it either way.