The sound generators of string and non-electronic keyboard instruments are nearly always coupled via the soundboard or organ case. Sympathetic resonance between two notes is strongest for pure intervals. With tempered intervals, this effect should be weaker, but in addition the coupling should "pull" the two frequencies away from the frequencies of the notes played separately.
People such as Werckmeister who developed tempering systems defined the intervals in terms of length ratios on the monochord. Nowadays, published data are given in cents, presumably on the assumption that the length ratios are an accurate indication of the frequency ratios. I'd like to know whether this assumption ought to be questioned.
A supplementary question would concern possible differences between tuning a harspichord or piano note-by-note with the aid of an electronic device, and doing it the difficult traditional way by listening to pairs of notes played together.