Even beyond fully 180-degree polarity and latency, there are frequency-dependent phase issues if you are blending two different examples of the same signal run through signal chains (amps, preamps, etc) with different EQ topologies and different settings- so maybe the bulk of the signal might be roughly phase-coherent when testing using a 180-degree phase-reverse test, but certain EQ topologies and settings might be manipulating the phase at only certain frequency bands that can result in funky cancellations (not necessarily bad subjectively, but also not easy to diagnose or fix) since EQ usually works by shifting phase in affected frequency regions.
Here's a pretty cool demo of this behavior that would be relevant pretty much anytime you are mixing two or more examples of the same source material run through different signal paths:
Personally I've found full-frequency parallel paths to be a pain to reconcile in the studio a number of times (trying to blend a clean signal with an "amp" type signal from a preamp or other chain, or blending two different amp or amp-sim signals for example and getting some strange holes in the midrange or other issues not present in either of the individual signals), to the point that I'd probably never trust it in a live environment without my own sound-person and lots of prior troubleshooting (and honestly I'm not convinced that it adds enough benefit in most cases). Biamping or crossover mixing is a different story though, as the two signals don't share as much of the same material to get cancelled out (still something that I wouldn't necessarily trust a typical soundperson to mix live, and would send a single line mix from my end if I was going to use that kind of signal split).