Interesting that test48, test70 (with reservations) and test80 do not look so hard to my solution. At least if the commenters in previous threads were right with examples of these tests. But mysterious test30 and test31 - I don't even imagine what structure they got.

Test70 is "randomly" solvable: with time distributed between 0.3 and 2.9 seconds (with avg = 1.1 sec) on my Intel i5-9400 (by "randomly" - I mean, of course, that could be 2.9 seconds which is TLE and would not pass)

And my suggestion: It would be nice to have a separate section on the site with parallel running of tests for such hard computational tasks. I mean: maybe someone also passes test 70, but fails test 30, for example. And the total percentage: passed N tests, failed tests with numbers K, L, M

Now I know what's been wrong with my solution (TLE30), but now TLE46

If you read SPb-MaxBuzz's article about test generation for the problem, then you can see tests rearranged intentionally for top of solutions at some time point in the past. You can think your solution is very special.