Well, the natural radiation of acoustic instruments is neither purely directional nor purely spherical – each instrument has an individual radiation pattern. Horn instruments have the narrowest radiation angle, at least at middle and higher frequencies. Triangles, cymbals, violins... have a rather wide dispersion even at higher frequencies, but radiation is still not exactly spheric. So a music reproduction through a perfect point source wouldn't exactly sound like the real thing, but in fact slightly too diffuse – and too uniform with respect to the different instrument characteristics. Nevertheless it would provide the closest impression to a live concert, provided that the room acoustics are adequate, i.e. rather on the dry side. But again: The radiation patterns of conventional «direct-radiating» speakers represent a rather unnatural dispersion, with lots of irregularities and most notably an increase of directionality toward high frequencies, whereas lower frequencies are allowed to radiate almost spherically.
Speaking of room acoustics: Normal living rooms, even if passably prepared for music listening, are a bad precondition for a lifelike reproduction of a large orchestra. The reverberation delay is much too short, there will always be the impression of a small room. Of course it helps that the recording already contains the large acoustics of a large concert hall, but reproducing the latter in a small room means adding inadequate, artificial reverberation to the original reverberation, i.e. reproducing concert-hall acoustics within living-room acoustics. Hence the increased reverberation produced by «omnidirectional» speakers – although actually providing a more even radiation characteristic – will naturally increase the «fake» effect. Therefore a countermeasure in the form of careful room treatment is indispensable to avoid a too diffuse sound experience. Apart from this problem «omnidirectionality» in fact leads to higher realism, simply because it implies a less frequency-dependent radiation characteristic. This makes for a relatively neutral indirect sound from all directions – necessary for a realistic reproduction of a concert-hall experience.
During my own radiation experiments I used acoustically «hidden» indirect radiators for the recreation of the missing higher frequencies within the reflected sound, I also used diffusors on top of conventional drivers, particularly dome tweeters, but also dynamic horn, piezo and ribbon tweeters. One of the latter designs with a Decca Kelly ribbon driver with a large horn and a special diffusor on top of it for 180° radiation provided a particularly realistic sound, so much so that a female friend of mine couldn't stand it. However, it was exactly the latter configuration which revealed a fundamental weakness in the reflector method: a certain hollowness and indirectness, the same effect that's also present in horn speakers, just to a lesser extent. Apparently the reflections within the reflector arrays (and the horns) leave their traces. At least smaller reflectors with smaller tweeters behave much less conspicuously in this respect, so they're a practicable compromise. The best results concerning this matter was provided by a piezo tweeter whose horn was shortened to a few millimeters, so it is virtually reduced to the compression chamber and some phase-correction channels. The small conical diffusor on top of it (which halfways recreates the horn function) is acoustically very inconspicuous.
A quite attractive commercial solution is Elac's 360° ribbon driver.
Unfortunately it can only be used at rather high frequencies without harmonic distortion getting out of hand – certainly the poor suspension plays a role here.
My own implementation of a 360° ribbon driver (only existing on paper) introduces an ultra-thin nylon or kevlar yarn (red) as super-elastic suspension. The downside is a complex form of the pole pieces which would be expensive to manufacture. In contrast to many other tweeter designs it also offers excellent ventilation enabling a large air volume and thus low resonance frequency and low inner reflections.
I haven't heard a fully convincing buyable 360° radiator so far. I've heard the walsh driver, but was rather unimpressed, so much so that I even can't remember the sound. I have heard the Magnat Plasma system...
...and there the three (or so) metal grids necessary for radio-wave suppression killed the sound. I haven't heard the MBL Radialstrahler...
...but expect them to be somewhat handicapped in terms of transient response due to the complicated drive principle and moreover the partial-vibrator design calling for resonance damping. I friend of mine called the sound «slightly artificial» (it was a different type than illustrated, though).
The most promising approach in my view (and substantiated by some imperfect prototypes) would be direct radiating 360° drivers of a rather simple and conventional principle: vertically radiating «dome» tweeters and squawkers with sort of a slightly curved cone in the form of a candle flame:
Unfortunately I'm not into speakers anymore. But if someone likes to get inspired by my ideas, more power to him/her!