This discussion is solely for purposes of discussing and arguing how they could work applied to the PF system, not whether or not they should be included in any version of PF. If you dislike the very premise of it, that is fine so long as you provide a decent criticism of the system to go along with your opinion.
I thought of this while reading the bounded accuracy thread, and also saw others propose similar later on in the thread. Sadly it got closed before I had an opportunity to post as well.
Sample system I have been thinking about. The bonuses on a roll can total no more than 10+level. Specifically though, the base skill or BAB would not be part of that limit.
An example attack would be a raging orc barbarian at level 1. BAB 1, 24 str for +7, masterwork weapon +1 height advantage +1. Total bonus to hit over BAB is +9. This character can receive only a max of +11 from any number of sources. This leaves the barbarian able to receive only +2 more on the roll no matter what modifiers get applied.
Things I see
Pros:
1. Players know at what point they do not have to worry about trying to squeeze out more and more bonuses. At a certain easily discernible point, more bonuses really do not help at all.
2. Balancing CR becomes (maybe) slightly quicker. Certain CR opponents/challenges are obviously impossible.
3. Leveling and choosing what you level (BAB or where to put skill points) matters more than finding the right splat book. A higher level character can succeed on a challenge that a lesser character truly cannot, no matter what weird race/trait/random splatbook #2345 a player finds.
3a. The goal is to narrow the gap between the utterly optimized and the just well optimized. NOT to bridge the gap between the optimized and unoptimized. A character not designed well to do something should not do that something almost as well as a character that focuses on that task.
3b. Just because of a glut of resources a character should not be so much better than another well optimized character that simply doesn't have access to those additional source books as to render an otherwise well optimized/designed character ineffective. It can be fun to find a way to get +30 to diplomacy at level 1. It should never be necessary or expected. This is not the same as a character being superfluous simply because the need for the task is already covered.
4. Continuing on from 3, this allows players to reallocate resources. There never needs to be a worry about an arms race with the GM (which the GM can always win). Players and GMs know what can be easy/medium/hard and what is impossible for a given level, no matter the degree of optimization.
5. Many spells give bonuses that already vastly shoot over these limitation. This could lower the power of some until higher levels when a character can make use of them, or invalidate some of them entirely.
Cons:
1. Could disincentive teamwork. In the above example, if a rogue provides flanking for the barbarian, he would become unable to benefit from a bard in the party using his inspire courage (for the attack at least, not on damage). Heavily optimized characters could instead get less benefit from having party members help.
2. Massive rebalancing could be necessary. Given the setup above BAB +10+level or Skill+10+level lets characters have decent success (often gauranteed) on any level apprpriate challenge I checked, but I have not had time to check them thoroughly.
3. Corollary to 3 above, finding just the right combination to make something work can become less effective. Delving into the resources and working out combinations is just fun for some people, myself included.
4. Many spells give bonuses that already vastly shoot over these limitation. This could lower the power of some until higher levels when a character can make use of them, or invalidate some of them entirely.
Most obvious low level spell is true strike. +20 to hit would max out a characters bonuses even if they have no bonus at all. Could be good or bad. Good: Useful for a character with a poor to hit to get a decent value. Bad: Becomes limited to ignoring concealment to anyone that has decently focused on attack.
Another idea was tying the bonus you can receive directly to BAB/skill ranks. The better you are at something, the better you could utilize things that make you better. Limit being 10+BAB or 10+Ranks.