Perhaps you could consider the facilitator question, "does this action make sense for your role?" as a time-saving step. It allows the role-holder that is being asked to take action (that is in attendance at the meeting) to quickly react and push for more clarity in the moment if a requested action doesn't make sense.
As an example, let's say a role with a purpose of "Training is delivered" is asked to consider taking the action of "training materials are reviewed for typos." In the meeting, they may say that the action simply doesn't make sense. It's outside the scope of their purpose. That will immediately trigger new tensions, such as "there is no role responsible for improving training materials" or, the originating requester realizes, "I have expectations that 'training is delivered' also means training materials are maintained." Those discoveries can generate new, immediate tensions to be processed in the meeting.
If the role-holder representing the "Training is delivered" purpose isn't at the meeting, what we do with those requests can evolve with what the circle needs. I believe the most efficient, streamlined approach is to capture the action and entrust the role-holder to process it later (without anyone assessing if it "makes sense" during the meeting). That role-holder will (hopefully) process the NA request exactly the same as any other request and say "this doesn't make sense for my role." They then have valid tensions to process... notifying the requester that they aren't doing that work or suggesting changes so that the purpose they fill isn't easily misinterpreted.
I think I would only suggest my circle invoke additional process (like asking the Lead Link to consider if an action "makes sense" for an absent role) if there's a consistently fuzzy purpose or a consistently absent member and it's causing a lot of processing/follow-up.
To be fair, all that frustrating, pedantic follow-up caused by an NA being requested in error is where the real benefit lies. If action is requested that doesn't fit, there is an unhelpful expectation somewhere that should be cleaned up. Finding and fixing that misalignment is where the real long-term solution is.