Calculating the exact sequence values (e.g. using BigDecimal in Java) and finally converting them to double fails with some of the random test cases.

Example case:

• signature = [4.0, 0.0, 8.0, 16.0, 9.0, 10.0, 16.0, 3.0, 6.0, 2.0, 19.0, 14.0, 5.0, 4.0, 4.0, 2.0, 7.0, 2.0, 7.0]
• n = 69

The exact sequence contains 18773643266614620, which can (just) be represented exactly as a double. The test checks for 1.8773643266614612E16, which is 8 less than the exact value.

Maybe random tests should be limited to a maximum value below 2^52.