Hi ewaostrowska, thank you for your time and effort maintaining this project.
I’d really appreciate it if you could take a look at this PR whenever you have a chance 🙏

Hi @juntae6942,
thanks a ton for opening the PR and digging into this!
Your change does fix the redundant schema resolutions and there are less random cache misses. Common types (such as Strings) won't be re-resolved over and over again. The provided tests are a great documentation of the problem.

There are however few cases that could be addressed in this area:

1. Order-sensitive annotations
   ctx = {@A, @B} vs {@B, @A} // NOT equal → extra resolution
   equals uses Arrays.equals(ctxAnnotations, …), which is order-sensitive. If callers produce the same effective set of annotations in different orders, there will still be missed cache hits.

2. JDK/internal annotation noise
   If one call site picks up a jdk./sun. annotation and another doesn’t, the arrays differ → not equal (even if hashCode tries to ignore them).
   hashCode tries to ignore sun./jdk., but equals still compares the raw arrays, would be nice to have these two aligned

3. Mutability hazard
   ctxAnnotations is used in equality but is mutable; changing it after insertion into a Set can corrupt lookups.
   ctxAnnotations participates in equality, and it’s an array stored on the object; if it’s mutated after insertion into a Set, lookups can break.

4. getClass() != o.getClass() disallows subclass equality.
   At the moment we don not inherit but it’s a behavioral change vs instanceof and would be good to have a final keyword on the AnnotatedType if we make a change. Currently I think it's better to leave it with instanceOf

5. Adding few tests to the cases above and toggling fields excluded from equality (name, resolveAsRef, etc.)

We really appreciate your contribution, and we’d be happy to see an updated PR with these extras. Are you interested in taking it further?

ewaostrowska Thanks for the detailed feedback! Yes, I'm happy to take this further. I'll look into the points you raised and update the PR.

Hi ewaostrowska,
I would be grateful if you could review my changes and let me know if the approach I took is appropriate.

hi @juntae6942, thanks for taking it further
I think the approach should fix cache fragmentation from the contextual fields, subclass equality, jdk annotations and addresses the mutability hazard which is awesome :) It does still need some work on aligning the hash and equal mismatch. Also some clean up would be necessary. Thanks for adding more tests as well