"For this Kata you need to write a function that takes a string of characters and returns the length, as an integer value, of longest alphanumeric palindrome that could be made by rearranging the characters."

I think this more clearly and concisely explains the goal.

1 is not a prime number. Therefore 1 cannot be considered a "factor", but 1 is a divisor of the base. Consider renaming the kata and objectives.

Since z/Z is the last letter to find, we only have to index to m/M.

1. It is unfortunate that the kata's requirement's changed to include pieces with holes. My original solution would have sufficed.
2. "There are no borders touching each others." Would the following shape be allowed (python)?

"\n".join(["++--+","++--+","| |","+---+"])

Can the attempt test cases remove the duplicates? See below.

Failed for n = 7: _
Failed for n = 25: _
...
Failed for n = 7: _
...
Failed for n = 25: _

If there are 100,000 tests, instead of having 999,990 pass and 10 fail, this just returns 10 fail.

Something is not correct when finally attempting this kata in python 3.8. My solution does not have any recursion. There is a single, exterior def, also not recursive, with a single print statement. That exterior def is called at-most twice, yet I see dozens of lines printed out per test. The sample tests work exactly as expected: I see at most two lines, and my solution passes the sample tests.

Now the input is not mutated.

I don't like that I mutated the input.