I was reading Wilhelm's post about Torchlight 2 being too late for the Mac market. Wilhelm points out that three spiritual successors to Diablo 2, Torchlight 2, Path of Exile, and Diablo 3 were released roughly around the same time. Yet of the three Torchlight 2, despite its developer carrying over a lot of good will after the cult success of Torchlight, more or less flopped into obscurity. The problem was that while Diablo 3 and Path of Exile correctly identified portions of the Diablo 2 formula people liked and disliked, Torchlight 2 didn't, and thus couldn't hold players.
Diablo 3 and Path of Exile figured out that players like to create various
builds, they like "lootsplosions," they like a good campaign, and some form of repeatable
"endgame" content. Both delivered this in various ways, but also identified the frustrations players had with Diablo 2 and mostly got rid of many of them. D3 addressed players wanting to try different builds by allowing free skill swapping. Path of Exile did it by creating a myriad of possibilities and making it relatively easy to reroll. Torchlight 2 on
the other hand effectively recreated the same "once you start you're stuck" philosophy of Diablo 2 that was more frustration than fun. It may have stayed true to the formula, but that formula was written over a decade ago.
Sure, TL2 allowed for various builds, but they weren't all that exciting and it required a lot of investment to try each. Sure, TL2 offered lootsplosions, but the RNG on gear was so high that a player could go ten or twenty levels with the same gear. Players like that periodic reward; it's fun to get a shiny new weapon every few levels. In both cases Runic misinterpreted what it was about those systems that players enjoyed, and instead tried to remain as close to the D2 formula as possible while only updating it with minor conveniences.
In a perfect world the developers of a game could read the minds of the players and from there figure out exactly what they wanted. Sadly in practice developers have to rely on metrics, player feedback, and the media. Interpreting that data is clearly an art. If your data is showing that players enjoy killing each other in the open world, the answer is not necessarily to reduce the number of safe areas and confine the map more. There is the additional risk that you get the wrong information; metrics can lie, not everyone fills out exit questionnaires, and direct community feedback risks hearing too much from "the vocal minority." With all that going on I don't blame devs for misinterpreting it all at times, but the market is less forgiving.