Electron-positron event shapes are among the oldest and most established observables for extracting the strong coupling constant. The most precise determinations based on the thrust and C-parameter distributions yielded values, however, that are currently in tension with the PDG world average. We therefore revisited the theoretical foundation of the event-shape fits, in particular their implementation of non-perturbative effects. Interestingly, we found that the combined effect of altering the renormalon subtraction scheme and perturbative scale parameters can lead to a few-percent-level impacts on the extracted alpha_s value, indicating a potentially important systematic theory uncertainty that should be accounted for.