GM-run events in MMOs do not scale well, and their compromises mean they pretty much have to be lousy.
MMO PvE content is designed to scale well, in whatever way it scales in your game. If you make a good quest or instance, you can have it running on hundreds of servers at very little additional cost. Having 10,000,000 people play WoW takes far less developer time than having 100,000 people play each of 100 MMOs. This is mass-production, with increasing returns to scale. Computers are good at copying the same thing for more people.
Developer-piloted NPCs do not scale well. They can be in only one place. If you want that NPC to appear in more than one zone or on more than one server, you need multiple humans to pilot them. Programmed and scripted events can be copied, but the human interaction portion cannot. If you want to give every server an equal experience, you are going to need a script, reducing the human element; you are going to need either staggered starts or a large team; and if you have staggered starts, you need some way to cope with people from the fifth server knowing the script already. Sadly, the end point must always be the same, because you cannot support divergent servers. Any human interaction affects only the storytelling, not the plot.
If the developers of EVE Online, Darkfall, or A Tale in the Desert (all sandbox-y games in increasing order of PvP hardcore-ness) want an event, they can go do it. There is a server or two. They do not need to worry about divergence or repeat performances. Single-server architecture is a grand thing. Similarly, players can run their own events and affect the world. Even if most of the “events” are from people who want to watch the world burn, you have a much larger upside potential.
(Page not found? Look harder.)