Ten years ago "mini-games" was an inescapable term in the industry. Every major proposal had to include some number of non-standard gameplay elements sprinkled throughout your AAA game design. Training mini-games in Fight Night, rail-shooter levels in Turok, gambling in Shenmue - mini-games were everywhere.
This always seemed like a poor mandate to me - as if we developers needed to pace out the main game experience because our players were too single-minded to vary their experience on their own. When playing a game I prefer to set my own pace - and switch to other activities periodically if my attention wanes. Pausing my latest exploration in FTL to peruse new messages on some forums lets me return to my experience refreshed.
But forcing breaks within the game means that the experience almost always suffers. If the mandatory break comes when I'm ready for a break then the game has effectively done nothing. If the mandatory break comes when I still want to focus on the primary activity - then I feel frustration. This can be mitigated somewhat by ensuring that the breaks are as "good" as possible - that they make sense within the context of the game, that I have some control (or at least warning) about when they will come up (and end!) and that they reinforce the main game rather than detract from it.
"Bad" breaks, on the other hand, are always failures regardless of when they come into view. A bad break makes it difficult for me to return to the main game, because it broke my interest. Network television commercials are often bad breaks - watching the same shows via internet streaming services has demonstrated that the shows work better without those breaks. Saturday Night Live was an early innovator in reducing the negative impact of commercial breaks - because some of the commercials were actually just pre-recorded segments of the show. That keeps the regular commercials more in tune with the show - turns some of the bad breaks into good breaks.
But there is another type of break - it is possible to provide players with a break when they wouldn't take one on their own, but after taking the break they appreciate the respite. These are very hard to gauge because they often focus on physical factors like fatigue or eye-strain, which don't reliably occur at specific points in the game experience. But when done well, these breaks provide the "bathroom break effect."
The great thing about a "bathroom break" is that it is all about removing you from your primary context, and providing you with a secondary, temporary context. You're not just avoiding one activity - you are actively focused on another, essential, activity.
The trick here is the word "essential". Inventory tetris isn't essential - neither are Diablo town-trips. (As evidenced by the amazing "dog" mechanic in Torchlight!) Perhaps the best example I've seen of a consistently essential activity was in the brutally hard DS game Henry Hatsworth in the puzzling adventure. This comes close to the dreaded "pat your head & rub your tummy" problem evident in games like The World Ends with You, but it manages to just barely skirt by on charm and integration.
The concept of breaks has faded in the past 3 years, largely because the market has become more focused on bite-sized games instead of huge epic time-sinks. But I think the lessons of good break design still hold - let your players manage their breaks as much as possible, by providing them with a variety of meaningful break options.