Assuming happiness is not an issue.
Cottages are not a long term advantage, the great people production could be scrapped in it's entirety and still give specialists a substantial advantage. A fully upgraded grassland town only supports itself, yes it has significant science output, but five commerce is only five commerce. A sanitation level farm on grassland provides 6 food, enough for its own maintenance and two more population. The biggest problem isn't really the agriculture civic, it's the combinations of happiness modifiers one has access to. With law mana and archmages, you can have massive farm cities producing astronomical science, commerce and culture while still having greater production than they would have as cottages, regardless of whether the tile is a plains or grassland tile. Not to say there is something particularly wrong with the order spell, or massive happiness modifiers from ancient forest covered elven power houses.
Cottages are only a long term positive if you can't increase your happiness far enough to make them obsolete. With sacrifice the weak and law mana, it would be insane to build anything but farms on flatlands in six cities, with or without agriculture at all. You don't even need the civic combos at all. You can skip aristocracy, caste system, agriculture and scholarship and still outproduce cottages with farms in every way.
In a massive empire, I'd probably end up skipping agriculture for mercantilism since I tend to like being at war with everyone anyways and wouldn't be able to keep the happiness up. To go for any other victory though? Great sage pops will help with the early tech rush and get you to archmages all the faster, and a three to six city empire with 500+ culture production, 200+ science output, massive industry more akin to an arete mine network, and still enough extra commerce to upgrade anything and everything you can think of... I strictly automate all workers to avoid cheesing the hell out of my own experience against the ai. If I ever micromanaged them...
The baseline for cottage effectiveness versus farms scales directly with happiness after you pass the maximum sustainable population with cottages, which isn't very high anyways, there's really no comparison once you add in ways to greatly extend the happiness limit.