Testing time is the larger problem.
I agree, that's why I suggested a total conversion mod, not an option. Automated tests (which I'm working on building more of) can test shared functionality and make sure it doesn't drift unexpectedly (i.e. by a code change adjusting the effective accuracy of a gun with particular stats), then each set of stats needs to be maintained separately, because if they have dependencies on each other, they also have opportunities to influence each other. There are too many points of potential disagreement for a single flag to toggle between modes in a sensical fashion, for example:
Maximum range of individual weapons:
Realism says ranges are very long, and weapon types aren't distinct, game logic says ranges are constrained by the game engine and weapon types have distinct ranges.
Scaling of hit rates compared to range.
Realism says hit rates decay linearly with distance, game logic probably wants a shallow hit rate decay at first followed by a steep drop.
Impact of increasing stats vs skills vs weapon type vs weapon modifications.
I.E. each of these contribute to accuracy, but probably in different proportions.
Handling of improvised firearms:
Realism says they're very inaccurate compared to factory weapons, game logic says they're comparable.
Existence of weapons:
I assume a more game-oriented fork would prune out a large number of weapons in the interests of simplification and reducing workload. Likewise a game-oriented fork might add things like freely constructible firearms that are on par with factory built guns, they would simply not exist in dda.
Trying to pack all of these into a single option is a recipe for the system being buggy and poorly tested. Some of these (e.g. hit rate decay with distance) would need to be toggled with a flag and implemented in the game engine, but others would need separate definitions for all supported weapons, and some weapons would be blacklisted or undefined instead.