I am glad you like the kata. :)

I collected several test maps that your solution does not agree with the test code. I used a more brute force (but slow) algrithm to calculate those maps again, which gave same answers to the test code. I guess the answers from the test code are more likely to be correct. When writing the brute force code, I found a way to improve test code speed. I updated the test code, but not adding any fix tests. Do you think I should add some 90 territory maps in the fix tests?

Not a bad idea.
How about to change it this way: Instead of true and false, the code should return how many conflicting cells. If zero, it means the net is valid. A line of 8 cubes [1,2,3,4,5,6,7,8] has 4 conflicting cells; while a T shape [1,2,3,10,18,26,34,42] has two. For those net not in one piece, return -1.

These won't need big changes in the test code.

Thank you for the reply. I finished the 1kyu kata, with a messy spaghetti code. That was the 1k I spent longest time to solve. really appreciate the great kata. thx

No, I am not worried about it. 8)
I mean if we change restrictions in test code here, losing some score can be a concern for the existing solution authors. I do not see other impacts.

I won't lose any score if my old solution becomes invalid due to a new test code update, I think.

I put a part of solution into the Test function in C. C should not have that concern above now.

I agree with you that there is no good way to limit hardcoded solutions.

However, I am not convinced yet the suggestion need be removed since test code can't police it. It implies the direction of solving, without missing the right experience. It likes a Klotski board game says you can only make 2D moves.

Done. Thank you.
Not good at C/C++. Please advise if anything else need modification. THX.

Thank you so much for the suggestions.
I modified C and C++ test codes.

I modified it as you suggested. Thank you!