It looks the fix(JDK-8280950) addressed the issue partially, I see the test (test_nextDoubleWithBoundDef1) using nextDouble(bound) still fails with the latest JDK nightly.

=======
  @TestCase
    @TestedStatement(source = "nextDouble(double)", value = {"The default implementation checks that bound is a " +
            "positive finite double. Then invokes nextDouble(), scaling the result so that the final result lies " +
            "between 0.0 (inclusive) and bound (exclusive)."})
    public void test_nextDoubleWithBoundDef1(@Doubles({Double.MIN_VALUE, 10.0, 1234455555.45, 10000000d}) double bound) {
        AtomicBoolean nextDoubleInvoked = new AtomicBoolean(false);
        RandomGenerator rg = new RandomGenerator() {
            @Override
            public double nextDouble() {
                nextDoubleInvoked.set(true);
                return 123456789.34d;
            }

            @Override
            public long nextLong() {
                return 0;
            }
        };
        double value = rg.nextDouble(bound);
        Assert.assertTrue(nextDoubleInvoked.get());
        Assert.assertTrue(value >= 0);
        Assert.assertTrue(value < bound);
    }

===========
 Possibly one more place to address it may be - public static double boundedNextDouble(RandomGenerator rng, double bound)