Trust is earned, not offered. And without the potential for trust to be revoked, it quickly becomes a status of nominal value.
Let's look at the closing system for an example.
- Originally, any individual 3K user could close or reopen any question instantly.
- When this generated complaints, the system was changed to require a vote each from 3 3K users to close or reopen a question.
- When this continued to generate complaints, the number of votes required was raised to 5.
- When it became clear that this was ineffective for the vast number of duplicate questions, duplicate-closing was restored to an instant-binding-vote system for gold badge holders.
So at #1, we have folks earning a level of trust by participating on the site in a manner helpful enough to earn at least 3,000 reputation points. Then in #2 and #3, that level of trust is lowered, after those who had earned it are perceived to be using it irresponsibly. Finally in #4, trust is in part restored for those who have met a much more stringent set of criteria. If that trust comes to be seen as misplaced, then we might see a corresponding reduction in privileges granted to folks meeting those criteria, or a tightening of the criteria, or some other mitigation strategy (revoking of privileges for individual users, etc.) OTOH, if that trust is seen as well-placed, then those privileges may be expanded, reflecting a greater level of earned trust.
Now consider another example: suggested edits...
- Suggested edits can be approved or rejected by two users who have earned 1K rep.
- When this was seen as confusing, the requirement was dropped to 1 reviewer with at least 2K rep.
- When this resulted in a huge amount of pointless edits clogging the queue, suggested edits were made to require two agreeing votes to approve or reject, or a complementary edit.
- When this still saw too many actively harmful edits being approved, the requirement was changed to three agreeing votes to approve or reject.
- In response to complaints that it was now near-impossible for a conscientious reviewer to reject trivial edits that left glaring problems in place, the criteria was changed again to require 3 agreeing votes, or a complementary edit for approval, or an alternate edit for rejection.
Again, we see an establishment of trust based on perceived criteria in #1, then either a reduction in trust or an increase in the criteria needed to earn that trust at each subsequent milestone.
These two examples illustrate something else... It is preferable to grant trust in response to demonstrated competence in the area where trust is being granted, rather than granting trust for unrelated achievements:
- Folks who answer a lot of questions on a given topic are generally pretty good at identifying questions that've already been answered on that topic.
- Folks who are able and willing to make edits themselves are generally pretty good at recognizing good/bad edits from others.
The actual answer to your question about imposing limits contradicting the community's preferences
We can't. Not directly, anyway. We can nudge things here and there to maintain balance when part of the community is disenfranchised - or becomes too powerful for the rest of the community to keep in check...
But we can't stop water from flowing downhill, at least not for very long. The best we can hope for is to direct it in a constructive direction.
Most of the time though, our job is just to make the system's idea of who it "trusts" roughly in line with the community's idea of who it trusts. And that's plenty hard enough.