How does the d100 system work?
At it's heart, the d100 system works on the idea of telling the player exactly what his percent chance of accomplishing a task is. You are given a binary success/fail with a 1-100% chance of success (or 1-95%). The Dark Heresy system elaborates on this by including an additional mechanic in degrees of success, which dictate how WELL you succeed or how BADLY you fail. These results are sometimes binary (e.g. on 3+ DoS, gain X; on 2+ DoF, inflict X damage to yourself), sometimes scaling (e.g. increase damage by 1 per DoS), and sometimes left to GM interpretation, depending on what you're trying to do. As I type that out, I feel like the system could seriously do with some consolidation for Degrees of Success and Failure, and what they mean. Maybe some sort of big table of ways to apply those degrees for players/GM within encounters would be nice.
The d100 system has characters testing against their own skills or attributes, thus defining the difficulty of any task by the character's own abilities. In essence, assuming no other detriments from context, characters faced with any task determine its difficulty entirely through their own skill. DH introduces contextual effects through difficulty penalties or bonuses, and other ways of increasing the chances of the roll. At it's core, however, is that player's starting attribute. Compare this a system like D&D where hits are all based on an external difficulty and skills are then added to the roll. Technically, yes, both ways of determining difficulty arrive at the same destination, but the route they take to get there determines how the players and GM perceive the mechanics of the game.
In particular, the modifiers for difficulty in Dark Heresy are fundementally unintuitive for players. This is because the player's conception of difficulty is resting within his character's own skill. The external difficulty of the task is secondary in the player's mind to what his own character is capable of doing, because that is how the base of the roll is determined. When you determine that a task is easy, it gets a +20 or +30. That still leads to your average human in DH failing the task 1/2 the time or 2/5 of the time. So what if you increase the task to where your average person can almost never fail. Why even roll, then? The only reason to make these kinds of rolls is to see how the Degrees of Success/Failure mechanic comes in, which I've already established isn't consistent across all the rolls that players make.
Consider this: you have two characters, one with WS 50 and one with WS 30 both trying to hit someone. Intuitively, and technically, it is harder for the WS 30 character to hit someone. However, the task of hitting someone, external to the characters, is equally difficult. Within the reality of the game, it is not more difficult for WS 30 to hit someone, but it is more difficult within the reality of the dice. Again, this leads to a disconnect between the fiction of the game and the game mechanic.
I mentioned "the average person" above, which DH defines as having around a 35 in all attributes. This average means that for any challenging task, completion only occurs 1/3rd of the time, with around a 2/5ths chance of 3+ degrees of failure. I'm not the first person to note how immersion-breaking this is. However, Dark Heresy briefly states that a roll should never be made unless it can result in something interesting happening. First of all, that rule flies out the window during combat, in which a miss is just a miss, and nothing more. Second, the binary nature of the rolling often leads to times in which a die roll could definitely lead to something interesting, but only for one of either success or failure. Although the "interesting" rule for rolling works well in theory, it is actually at odds with the mindset of the system. The binary dice result leads to players striving for success, adding on modifiers. The GM, in turn, adds on penalties. These bonuses and penalties are not about how interesting an action will be; they are about achieving the binary result. The mindset of the GM and Players is guided toward thinking about success/failure, NOT how interesting the roll is.
Of course, it's simple enough for players and GMs to work around these problems, but I'm arguing that the d100 system by its nature guides players toward these issues by nature of its design and effect.
Something else to mention when it comes to using the d100 system at different power levels is the aforementioned self-determined difficulty. Given that the difficulty of a task is determined by a character's own ability first, it becomes impossible to balance characters of widely differing power levels within the system. The modifiers for "Easy" and "Hellish" are not actually given any grounding within the system. Consider the characters with different WS values. Even though the core difficulty of a shot is based on character attributes, the modifier is based on the factual difficulty. Players and GM have to somehow simultaneously grasp these two different philosophical metrics of how difficult something is. Should picking up a car that is easy for a space marine be difficult for a human? What happens when you reach feats that are literally beyond the ability of a human, but easily within the ability of an alien or space marine? If you try just saying "a human can't do this but the space marine can attempt it", any difficulty modifier you apply is suddenly based on the character rather than the factual difficulty of the action. And that's where the contradiction occurs. The game's design cannot simultaneously handle these things without violating the assumptions of itself.
So, to sum up:
The d100 system bases the core difficulty of an action on a character's factual internal ability
The d100 system modifies that difficulty based on factual external reality for the character
The d100 system runs on a binary Pass/Fail Mechanic
The d100 system modifies the Pass/Fail Mechanic with an inconsistent Degrees of Success/Failure System
Players and GMs are naturally inclined to first view the difficulty of an action by their character's ability, and this in turn conflicts with the ability to determine external difficulty
The d100 system sees the average person failing any challenges 2/3 of the time
The d100 system has a rule to only roll when a task is interesting, but...
The Pass/Fail Mechanic, inconsistent DoS/DoF mechanic, and the system of modifiers all guide players toward focusing on conflict resolution rather than the narrative weight of a task, leading to the above rule getting ignored and not working within the game's framework
The system is unable to simultaneously handle having creatures with greatly differing abilities due to the system for difficulty modifiers not being able to ground itself without violating the assumption that it is based on external difficulty.
Or, to really sum up:
The d100 system kind of sucks, the DoS/DoF system is poorly implemented, and the system can NOT handle having large power differentials in characters/creatures.