I saw this thread a fes days ago, but I didn't have time to read and reply to it at the time... but I do now.
Building a solid game foundation is never an easy thing to do. All of the points you brought up across the article were valid for all types of games from all tpes of creators;
Linear vs Free Roaming
There are certain aspects of each that can be considered a pro and con depending on the type of game you are making. It seems to be these days that every developer thinks that a "free roaming world" is somehow automatically better than a "linear" type of world.
This really depends on the type of game, the storyline, as to which option would be better. I enjoy both types of game modes, but only if it is suited to the game. I have played games where I thought "This game would have been way more fun/better/longer if it was a free roaming world", as well as "This game would hav been way more emerssive if it was linear and I didn't spend 1,000,000 hours trying to find the next mission/storyline/character".
Casual Fun vs Stupidly Hard
I like to die. I like to play a game and die die die die and die some more. I don't like to spend $100 on a new game, and finish it in 2 days without a challenge. 'Ninja Gaiden'... I died more playing that game than I did playing 'Another World'. I think that a lot of people just get too frustrated when playing games that are too hard... so there needs to be a difficulty adjustment ranging from "cake walk" to "fucking impossible bullshit wtf"... then make it a dynamic self-adjusting difficulty so if you die 5 times trying a section of the game, it automatically drops the difficulty down a slight amount or at least prompts the player with something like "you sure do die a lot, would you like to make the game a little easier?".
The Draw Card
Picking the "hero" and his/her "purpose" is always something that is gong to have ranged appeal... unless you make a female elf magicial nano techno steam punk bondage jetpack transformer, with a machine gun that shoots heat seaking magic missles and flaming skulls. But then that would look kinda funny if it were say, a mystery/adventure game set in the 50's. O.o
There is no winner here. Like newt said, you are aiming for a niche audience and you have to figure out what that audience is, by doing a little brainstorming/market research. Write down the types of people that play games and what they do for work/education. Then work out which of those people you want to reach the most. Then figure out what types of games those people like the most, and make your game based on that.
There is no right or wrong way of going about this, as no matter what you decide on, SOMEBODY will always like it and SOMEBODY will always hate it.
Conclusion
There is no spoon. I guess these questions have risen many many times before, and i seems like the "meaning of life"... nobody really knows.
Why are there 100,000 different kinds of cars to drive? Why are there billions of different clothing combinations? Why do some people not like chocolate?
There is no magical answer. There isn't even a half magical maybe kinda guess sorta. Make the game the way YOU want to make it, just avoid anything that will take the attention away from your main aim of the game, be it storyline driven, experience/collect things driven, statistics driven, action driven, etc. If you ar aiming for major storyline, avoid too much difficulty... going for a "collect things" style? Definately go free roaming... Action game? Make it hard and linear!
That's about all I can really say about all of that. O.o
Good luck, hopefully everyone's input so far has been helpful in some way.
~Sol