@Sasha.Brodsky,
This is a very interesting question. Unfortunately, I think the short answer is no.
However, the long answer is as follows.
The difficulty is that most merit function operands can only target a single value. What you are proposing is to have an operand that, for example, selects the 1st and 2nd surface of a singlet, and then constrains those surfaces to match a catalog that defines acceptable combinations of the 1st and 2nd surfaces in both a forward or reverse orientation. Now that operand is trying to optimize to any number of possible solutions that have to represent minima or roots in the merit function space.
I wrote a paper a while ago that describes how you can do this, at least for shape factor, and implement it either with a ZPLM or with a script that writes the merit function for you (a much faster computational solution). Shape factor only represents a catalog of two as ±1 count as a single solution and 0 is the other option. Different shape factors like for meniscus lenses simply represent additional roots in the equation that defines the merit function space but the problem is that there is not a consistent value for meniscus or non-equi lenses.
Here is a link to my paper.