The note in [conv.qual]/3,

[Note: Given similar types T1 and T2, this construction ensures that both can be converted to the cv-combined type of T1 and T2. — end note]

is close to being vacuous: A can be converted to B precisely when cvc(A,B)=B, so saying that T1 can be converted to cvc(T1,T2) is saying that cvc(T1,cvc(T1,T2))=cvc(T1,T2). It seems obvious that an operation like cv-combination should be idempotent in this sense (here, the unary function cvc(T1,·)); it is also reasonable that it should be symmetric (the reference to P_i¹ at the end of /3.2 damages that, but only in cases that are irrelevant anyway). If any note is needed, it is one that reassures the reader that those two properties hold (in all the cases that matter).