One example: https://codeforces.me/contest/2268/submission/392218401. It simply adds to an O(n^2 log n) algorithm a pruning that exits if the value is worse than the optimal answer for an interval of length 2, and it does not change its complexity. It has already been hacked after the contest, but it easily passed the system tests and in-contest hacks.
As we all know, CF Div.1/2 only has an in-contest hacking phase and no post-contest open hacking. The flaw of this format is that most contestants spend their time during the contest solving problems and do not pay attention to hacking; moreover, you can only hack within the same room, and only if you have solved the corresponding problem. This makes harder problems very difficult to hack, even if their algorithms are completely bogus, because there are already very few contestants in a room who solved the hard problem, and even fewer who, after solving it, would carefully read others' code. As a result, such code can also pass hacking and system tests. However, even under such circumstances, there were still quite a few successful hacks, which fully demonstrates how weak the system tests are.
If you can get a problem accepted by just writing a brute force, what is the point of CF contests? We did not write this brute force precisely because we believed that CF's system tests and hacks could stop incorrect solutions, but this contest betrayed our trust. Therefore, I call for the immediate creation of strong test data and a second system test together with the in-contest and post-contest hacks; otherwise, the fairness of CF contests will be difficult to guarantee.
