Some thoughts:
Olson's examples are worst case scenarios in many regards, using a source with extremely low directivity and placing it in non-ideal structures with the worst possible placement. If your source is meaningfully directional at the frequency which the baffle interactions occur you lessen them, because less sound reaches the edge and is reflected back to interfere with the direct sound.
Diffraction is not inherently bad, and can be used to extend the low-frequency directivity of a speaker or ameliorate other directivity aberrations. Since the frequency response aberrations introduced by edge diffraction primarily exist on-axis and dissappear as you move off-axis. This means that a peak in on-axis response due to diffraction doesn't exist off-axis (unlike, say, a resonance). If my cabinet introduces a 2dB peak on-axis at some frequency range due to diffraction and I equalize this down or fix it in the crossover then the off-axis response becomes 2dB lower, effectively making the loudspeaker more directional! A similar concept can be utilized to smooth out irregularities in waveguided tweeters or midranges. If my waveguide has a slight narrowing in the beamwidth at some frequency I can try to introduce a small diffraction dip at this frequency, which after equalization effectively rises the response off-axis.
Note that the above is not simple to do, and requires adequate simulation tools (e.g. AKABAK) or significant trial and error. If you want a response that is easy to work with and highly predictable, a sphere is of course ideal, just be aware that its DI will approach 0dB at a higher frequency than a similarly sized rectangular box. There's honestly a lot more that can be written about diffraction, but I'll end it here and say that I think it gets a bit of a bad rep when it is actually just another interesting tool in a designer's toolkit.
As an anecdotal example: when I was designing my midrange waveguide for a speaker project I am working on, I found that a 350mm wide cabinet without roundovers yielded a smoother directivity than a 370mm cabinet with 30mm roundovers! The effects were marginal, but they were there.