Chronologically:
Diablo 2 - I expected dark and atmospheric roguelike lite set in pseudo-medieval somewhat Manicheist setting, I got PGP-13 MMO lite with MMO mechanics (including lack of proper saves, even on exit), generic fantasy setting and unimpressive mobs. First and last game I ever kicked myself for buying, though it eventually got just barely fun enough in coop.
Unreal 2 - I expected insanely atmospheric and groundbreaking FPS like its predecessor - I got:
made for X-pox.
Oh, and derpy spaceorcs posing as skaarj.
Oblivion - well, duh.
I still can't decide if I hate U2 or TES IV more, but when I think about either I still see red and something still boils deep within me.
Arcanum does still provide a rather impressive number of solutions for the quest. You can kill the thieves, pickpocket the key to the gate, pay the toll, use persuasion or complete the thieves' quest.
Actually, that's about the standard for, say, Morrowind.
If you need item A from character B who is not wearing said item, the option to kill and loot and the option to pickpocket are pretty much given due to basic underlying mechanics of the game, plus you often get to persuade character by whatever means available, including bribery.
This is how a well-designed quest system works. It doesn't force the player to make a choice before he even knows what his choices are, much less whether he's actually making a choice.
This. It's ok to make the thieves suspicious but this shouldn't be a complete lock-out, not on this stage of the quest - you could've been meaning to help the constable, but you could just as well have been bullshitting him or simply be open to any potentially worthwhile alternatives.
A game that allows you to create any character you want should make possibly few assumptions about that character, else it's bad design. Coincidentally, making possibly few assumptions simplifies quest design while still allowing for meaningful consequences - sure making thieves wary of your meeting with constable would be a neat bonus, but it isn't necessary for the quest to work - letting your deeds speak for you would be enough.
Great quest design isn't recognizable by a spiderweb of extensive scripting. It's recognizable by making great use of mechanics already in place, and making possibly few assumptions to inevitably be broken when player decides to be creative.
