I've noticed many situations where a researcher will put their spouses as co-authors in a paper. This is commonly seen in situations where both spouses are in the same field. Valid collaboration between spouses is possible (and exist).
The question is not meant to attack spousal collaborations in academia. Simply put, this is just one possible example where a conflict of interest might arise (the same question can be placed in situations of parent-child and parent-sibling collaborations with no significant difference).
However, where does one draw the line in such a situation? What is stopping them from putting each other in every paper they write, effectively "doubling" their academic output? Whether we like it or not, publication metrics are used in fellowships, grants, awards, and many other situations in academia. In order for it to remain an objective metric requires the integrity of authorship to be upheld. When a PI is choosing authorship, they are responsible for not only themselves, but those in their own group, as well as their department and institution.
The situation with spouses (and other familial relationships) is somewhat unique in that the barrier for co-author contribution is far more easily passed, compared to two regular collaborating PIs. Not only are they often more readily available for collaborations, there is also a possibility of subconscious bias when evaluating the contributions of a significant other. There is also a greater possibility such a system can be gamed. For example, could such a possibility arise where a contribution which would otherwise be recognized in an acknowledgement be replaced by co-authorship due to a slight bias?
The issue is not whether or not spouses, family members or any other person should be allowed to be a co-author when they have made a valid contribution. The problem lies in whether or not the judgement of valid contribution can be made fairly without a conflict of interest by someone in that situation. Being aware of this, is it ethical to still make that decision? Is there a way to be unbiased without complete recusal?