JDK-8157000 introduced a mechanism that caused some overrides to be displayed differently from others. The implementation of that mechanism seems to be confusing for authors, who don't have a clear mental model to built expectation on.

* One example of circumventing that mechanism can be seen in jdk.incubator.vector:

    // Note: A surprising behavior in javadoc
    // sometimes makes a lone /** {@inheritDoc} */
    // comment drop the method altogether,
    // apparently if the method mentions an
    // parameter or return type of Vector<Double>
    // instead of Vector<E> as originally specified.
    // Adding an empty HTML fragment appears to
    // nudge javadoc into providing the desired
    // inherited documentation. We use the HTML
    // comment <!--workaround--> for this.
 
* An attempt to update the mechanism for covariant returns can be seen in JDK-8219147.

The bottom line is, the mechanism introduced in JDK-8157000 must be reviewed and clearly specified in the Documentation Comment Specification for the Standard Doclet.