"seeing the differences in the result"

This just means that you are testing against a very narrow output. It's actually pretty common to run across tests that don't even check for the likely failure cases, because the developer(s) don't actually understand the algorithm.

A common example is prime factorisation, most nontrivial factorisation algorithms (Pollard rho, elliptic curves), don't guarantee producing a prime factor they simply tend to produce them because they prioritize small factors. Programmers see that their function produces primes for the one or 2 test cases (out of say 2^64) and assume that it works. It generally does, but when it doesn't you get incorrect results (often undetectably) which poisons all the rest of your calculations.