For the gunshot example specifically, my intuition is that it should be possible to fool someone into thinking a gun was fired if the signal chain from capture to speaker is designed with this goal in mind.
Basically the opposite of a movie where viewers want to be able to comfortably hear dialogue and gunshots.
Gunpowder has about 3000 joules of energy per gram. A shotgun cartridge might be 30 grams. A decent chunk of that energy ends up as the audible shockwave. That shockwave comes out of the muzzle in about 1 millisecond.
Therefore, a speaker system would be needing an instantaneous output power of 90 megawatts to replicate it. 90 megawatts is a small powerstation if you wanted that much energy continuously. Typical large home speaker systems might go up to say 100 watts - or 900,000 times less power than needed.
(this analysis glosses over lots of details, but I believe the conclusion to remain the same even if you take those details into account).
Actually I think recording the sounds with a good dynamic range is probably the harder problem, but I'm pretty sure with some post production that would be possible as well.