Thank you for the clarifications.
We are using regex internally, and right now, it is only accepting 0-9,a-z,A-Z, underscore, hyphen and whitespace, in the key between the square brackets.
I see that this is not documented at this time.
We are open to adding colon as accepted text in the key, but we would have to expand our unit tests first to make sure that there are no regressions.
In the meantime, are you aware of any other characters that might be in your keys? such as & ; % $ etc.?