Thank you for the PR! I like the implementation. Clear, well-written, and tests. Nice!

I have some hesitations that I need to ponder:

• Against: I don't believe this is a common need for most applications.
• Against: It expands public API of Pundit, which needs to be carefully considered and maintained.
• For: Having an official reset allows people using Pundit to trust this to be future-proof. That has value.

Before this would be merged, assuming it will be merged, I think some changes should be made:

• Documentation change mentions "accounts", but we're dealing with user in Pundit. The hypothetical method should probably be switch_user to keep things simple.
• I believe the method name should be pundit_reset, to follow the convention.
• I'm not certain of this, but we should probably clear the flags too.
  • @_pundit_policy_authorized
  • @_pundit_policy_scoped

@Burgestrand Thank you for your thoughtful feedback! I appreciate your comments and am glad the implementation resonates well with you.

Regarding the concerns:
Against: Common need for most applications: I understand that this might not be a widely requested feature. However, I believe there are certain cases where having a reliable reset mechanism would greatly benefit applications with complex authorization flows. That said, I see your point and think it’s important to weigh the demand versus maintenance cost.

Against: Expands public API: You’re absolutely right :)

For: Future-proofing: Agreed! One of the key advantages of adding this feature is precisely the long-term value it provides for those who rely on Pundit. It gives them confidence that their systems will remain stable and maintainable as the library evolves.

Now, for the specific changes:

• Documentation and naming: Excellent catch! I agree that the documentation should refer to "users" rather than "accounts" to maintain consistency within Pundit’s terminology. I'll update the method name to switch_user as well—it keeps things intuitive.
• Method naming convention: I’ve renamed the method to pundit_reset! to follow Ruby conventions for methods that perform a significant change or could have side effects. Since the reset operation clears important state, I think the bang method makes sense here.
• Flag clearing: I see your point about potentially clearing the flags. I’ll investigate this further to ensure we aren’t leaving any unnecessary state behind after a reset. I'll review both @_pundit_policy_authorized and @_pundit_policy_scoped and make the necessary adjustments if needed.

Thanks again for the detailed review! I’ll proceed with the suggested changes and address the concerns you’ve raised.