One of the reasons: if you want people to buy your next thing. (I was fairly highly ranked online in one of the Ace Combat fighter pilot franchise games, and had built a stable of planes in another one. But when they shut down the servers for the previous installment, after introducing a new one, I found I couldn't get interested in investing in the new one, if it was only going to get ripped out from under me like the previous.)
The only place where I see that not happening is the future where game are 100% streamed and your computers is just a screen + controller (stadia 2.0). Then they'll have to settle for knock off clones of your game.
This is always one of the biggest issues with trying to make a private server. Ideally, we should be proactively capturing game data packets and archiving them for every game that has an online component, to ensure that somebody in the future will have that available if they're interested in developing a private server for whatever game.
Like others said - do you kill your product and probably company? Open source it :)
Over time the modding community has leverage this high level by building unreal native tools. The net result has been the same sort of cross tool use we saw with Gamebryo. Back with Gamebryo the Civilization modding shared tools with sid meier's railroads and even early skyrim and fallout modding.
Now the Unreal tools mean mods can be done in blueprint and should not break after every update like C# based mods. Likewise the common pak format used by unreal games (common because consoles prefer the pak for faster loading), means mods can replace files without overwriting files. Aka, mod conflicts become a reference overwrite issue and not a file overwrite issue.
As a developer of an unreal based game its pretty nice to know modding is no longer locked out. You still need/should provide the community with a modkit, but even those the community can generate on their own if they care.