KeyStore.load() javadoc (part) is:
    /**
     * <p>In order to create an empty keystore, or if the keystore cannot
     * be initialized from a stream, pass {@code null}
     * as the {@code stream} argument.
     *
     * <p> Note that if this keystore has already been loaded, it is
     * reinitialized and loaded again from the given input stream.
     *
     * @param stream the input stream from which the keystore is loaded,
     * or {@code null}
   ...
     */
With JDK-8062552 behavior of this method actually changed: prior JDK-8062552 then "null" was passed to load method for already loaded KeyStroe nothing happened (which is a believe incorrect as javadoc states that null will result in new keystore creation), after JDK-8062552 - new keystore is created.
This behavioral change cause user confusion. E.g. JDK-8137256 and JDK-8157220 that user created one after another insisting that it is a bug.
I think it would be better to empathize that KeyStore.load re-initialize and load the contents of the given input stream only when the input stream is non-null. And passing null will always result in empty keystore creation.