ok I am at 4kyu and I had a very hard time solving this problem.
Thinking: I am not that good of a programmer.

okay, the provided test cases seem to be "semi-optimal", since the random tests tend to fail. I didn't change the code compared to the version in my first reply, but after a few times pressing attempt&wait, the kata was completed.

Pressing "Test" passes the tests.
Pressing "Attempt" solves many tests, even many data-sets within the random_tests.
However, the random_tests fail after many sets.

STD Err:
main.cpp:63:34: warning: comparison of integers of different signs: 'size_t' (aka 'unsigned long') and 'int' [-Wsign-compare]
for (size_t index = 0; index < lhsSize && areEqual; ++index) {
~~~~~ ^ ~~~~~~~
main.cpp:155:50: warning: 'sizeof (output)' will return the size of the pointer, not the array itself [-Wsizeof-pointer-div]
assertArraysEquals(output, sizeof(output)/sizeof(int), output_gt, sizeof(output)/sizeof(int));
~~~~~~~~~~~~~~^
main.cpp:155:89: warning: 'sizeof (output)' will return the size of the pointer, not the array itself [-Wsizeof-pointer-div]
assertArraysEquals(output, sizeof(output)/sizeof(int), output_gt, sizeof(output)/sizeof(int));
~~~~~~~~~~~~~~^
3 warnings generated.

Max Buffer Size Reached (1.5 MiB)

-> now, did I screw up something, or are the random_tests crashing from time to time?
-> PLUS: The random test input arrays are NOT sorted as stated in the instructions.

WOOOOSH

Let me rephrase it, then: Besides, C++ is just C, but with classes, isn't it? BIG WINK EMOJI ;););)

I don't agree with you.

Modern C++ is not just C with classes.
Completely different constructions are used to solve problems.

Duplicate of a suggestion raised already 2 years ago. Author explicitly stated that they want the function signature this way.

Besides, C++ is just C, but with classes, isn't it?

Why is a function signature in с-style given for C++ kata?

void check_is_contained(int **mat1, const int mat2[3][3],const int nrows, const int ncols, int &row, int &col)

I propose changing to c++ style

std::pair<int, int> check_is_contained(std::vector<std::vector<int>> mat1, std::vector<std::vector<int>> mat2)
{
  // solution
  return { row, col }; // or tuple
}

https://www.codewars.com/kata/reviews/5a4b7c31b3f0a15f7200000c/groups/5ea10ceece3ee600010f4bb0

https://www.codewars.com/kata/reviews/5a4b7c31b3f0a15f7200000c/groups/5ea0929b02cf740001c01a50

https://www.codewars.com/kata/reviews/5a4b7c31b3f0a15f7200000c/groups/5ea0922502cf740001c01513

https://www.codewars.com/kata/reviews/5a4b7c31b3f0a15f7200000c/groups/5ea08dd42c977a0001dfcbcc

https://www.codewars.com/kata/reviews/5a4b7c31b3f0a15f7200000c/groups/5ea07aebc0c41f0001473fb9

Nvm misunderstood the task.

since you're not given the number of mines on neighbours surrounding R, I guess not?