False negative with diamond syntax

[cpovirk]

==> Foo.java <==
import org.jspecify.annotations.NonNull;
import org.jspecify.annotations.NullMarked;

@NullMarked
class Foo<V extends @NonNull Object> extends FooParent<V> {}

==> FooParent.java <==
import org.jspecify.annotations.NullMarked;
import org.jspecify.annotations.Nullable;

@NullMarked
class FooParent<V extends @Nullable Object> {}

==> TestClass.java <==
import org.jspecify.annotations.NullMarked;
import org.jspecify.annotations.Nullable;

@NullMarked
class TestClass {
  static <V extends @Nullable Object> FooParent<V> shouldFailButDoesNot() {
    return new Foo<>();
  }
}

(I included the redundant extends @NonNull Object above so that I could make a fairer comparison to typetools, which AFAIK doesn't yet recognize the effect of @NullMarked on type-parameter declarations. The issue reproduces even without that (and also with just extends Object).)

checker-framework-3.42.0-eisop4 $ checker/bin/javac -cp $HOME/.m2/repository/org/jspecify/jspecify/1.0.0/jspecify-1.0.0.jar -processor nullness Foo.java FooParent.java TestClass.java -AconservativeUninferredTypeArguments
Note: The Checker Framework is tested with JDK 8, 11, 17, and 21. You are using version 22.

This happens only when using the diamond operator: If we instead specify an explicit type argument V, we get an error (even without -AconservativeUninferredTypeArguments):

TestClass.java:7: error: [type.argument.type.incompatible] incompatible type argument for type parameter V of Foo.
    return new Foo<V>();
                   ^
  found   : V extends @Nullable Object
  required: @NonNull Object
1 error

typetools likewise gets this wrong at 3.42.0. It gets it right by 3.45.0 (I didn't check versions in between), I assume as a result of the type-inference revamp:

TestClass.java:7: error: [type.arguments.not.inferred] Could not infer type arguments for Foo constructor
    return new Foo<>();
           ^
  unsatisfiable constraint: V extends @Initialized @Nullable Object <: @Initialized @NonNull Object
1 error

My expectation would be that eisop would start working once it picks up the type-inference changes. I am filing this issue anyway just in case it's of interest. (I at least found it surprising that -AconservativeUninferredTypeArguments didn't help, though again, that's the case in typetools, too.)