If having epsilon much less sex would make us less happy, then having epsilon much more sex would make us happier, unless we are at exactly the local optimum, which sounds unlikely a priori.

Human preference doesn't consider gains and losses symmetrically. We are dramatically more sensitive to losses than to gains.

By some popular metrics, sex acts are discrete (if not always discreet), not continuous: you can't have epsilon more sex acts; you can have one more sex act.

If having epsilon much less sex would make us less happy, then having epsilon much more sex would make us happier, unless we are at exactly the local optimum, which sounds unlikely a priori.

Asr is right, but it actually goes so much further than this.... You have also entirely failed to account for opportunity cost. You aren't just adding some epsilon of sex, you're subtracting from some other area of life to have more time for sex.

And these are just the abstract, theoretical problems with your suggestion. The real-world practical problems of adding more sex are enormous... There are serious bottlenecks to sex. Both partners have to be in the mood or the act has potentially serious negative utility. Synchronizing desire takes a lot of time and effort. (It takes much more than a minute of romance time to yield a minute of sex.)

1. It seems pretty plausible that we might be somewhere near the local optimum, or more precisely that most couples are somewhere near their optimum. (I see no reason to think that different couples' optima are in the same place, or that everyone's on the same side of their optimum.)

2. You're now talking about epsilons, whereas your thought experiment involved getting couples to not have any sex at all for six months. That's a pretty big epsilon. [EDITED to add, for clarity: If instead you asked these couples to skip sex on 10% of candidate occasions in the next six months, or to try to have 10% more than they'd otherwise have had, I don't think it's at all obvious what to expect in either case.]

3. So far as I can see, no one is claiming that there is no causal relationship between sex and happiness; only that the observed correlation between sex and happiness may be largely the result of a different combination of causal relationships. In particular, it seems somewhat plausible that the following might all be true. (a) Any given person, in any given relationship situation, has a rough optimum level of sexual activity. The same goes for any particular couple. (b) Deviating far from that level tends to make people less happy. (c) Being in a stable, well-functioning romantic relationship generally tends to make people happier, to increase their optimum amount of sex, and to make them have more sex. (d) People and couples tend to adjust to roughly their optimum amount of sex. If these are true, then there will be a correlation between sex and happiness because of (b), but for any given person or couple to increase the amount of sex they have might be counterproductive because of (d). And if these are true, then your thought experiment will have the outcome you predict. (Note: Everything about "couples" in this paragraph should be taken also to cover larger groups of people who have sex with one another.)