What is stopping them from putting each other in every paper they write, effectively "doubling" their academic output?
Reputation. I think a lot of people will be skeptical about this very question, and that in itself is a drawback. People may wonder, when siblings or spouses or parents and children publish together, whether they were actually both pulling their own weight. I would not be surprised if many readers assume that more of the contribution came from one than the other (based on which name they recognized, who had a more prominent position, or other factors) while knowing nothing about how the work for that specific paper was conducted.
I have submitted to conferences with someone who happens to share my last name and is of the opposite gender. I have no doubt that some people would speculate that we were married, and I wondered at the time whether that would change their perceptions of the work. (The reviews were double-blind, so I did not worry about that for acceptance to the conference.)
To avoid untoward assumptions about whether it's all just the work of one partner, I imagine that both people should strive to publish separately as well, or with other collaborators, and/or they should have so much output that it would not be feasible for one person to secretly do all the work.
The situation with spouses is somewhat unique in that the barrier for co-author contribution is far more easily passed, compared to two regular collaborating PIs.
Sadly I cannot remember who this was, but there was a husband-and-wife team who frequently published together, and when one died, the other kept their names on papers, posthumously, for years until that part of the research agenda that had been in the pipeline was complete. My sense was that they worked so closely in developing the ideas over the years--as well as applying for funding and conducting the research--that the deceased spouse's intellectual contribution remained clear far after their death.
Finally, there was an interesting little paper about this in an answer to the question Kimball linked to. The authors, four unrelated individuals all named "Goodman," look into people who publish with the same last names, and they also point out that there would presumably not be the same "et al." penalty in this case, since both/all authors "sound like" the first author whenever it is mentioned.