When you don't get something done, when you fail or make a mistake, it can be tempting to try to make up for it. You can work extra hard tomorrow for the work you didn't get done today, or do something nice to make up for being crappy to a friend. It seems like such a good idea because it gives you the chance to take back a mistake, to salvage a bad situation. There's also a component of justice; you don't just get away with doing the wrong thing, you have to pay for it in extra effort later.
But does it actually work? I recently tried to employ some atonement after falling behind on my writing. I decided that the best thing to do was just write lots of new posts to catch up. I said at the time that I thought it would be a good disincentive, but mostly what I got to show for it was more failures later on. I've tried to do similar things with my prototypes before as well, where I fall behind but tell myself I'll catch up the next day. The results were... unimpressive.
A study in 2000 showed that disincentives can have a paradoxical effect; daycare centres that started fining parents for picking up their children late actually found it led to more lateness, not less. One of the mechanisms proposed by the authors is that this replaced a relatively costly violation of a social norm with a product you can purchase (the fine). Similarly, when you have some way to atone instead of paying the high psychological cost of failure, atonement can become more attractive than success.
I found that the atonement for falling behind on writing put me in a place where further falling behind no longer seemed like a failure, but just a continuation of the atonement process. What was previously a useful signal to tell me that I'd made a mistake just faded into repetitive background noise. The feedback it could have given me was wasted. Making up for prototypes had a similar effect; instead of choosing between do-it-and-succeed or don't-do-it-and-fail, I had a disastrous third option: don't-do-it-and-don't-fail.
It's not even really clear that atonement is based on a reasonable model. In the general case, you can't expect later actions to make up for earlier ones. Moving out of the way of an oncoming train has a very finite window of effectiveness; even the most extreme movement won't help once the train has hit you. An extreme example, perhaps, but really there are quite a lot of irreversible processes, things you can't take back or make up for. I think reversible processes are comparatively rare, and mostly exist because we go out of our way to create them.
In a sense, the quest for atonement is a quest for time travel. It would be wonderful to pull out your time machine whenever you make a mistake, but not physically plausible. When you make up for something, you're not doing the same thing but later, you're doing a new and different thing. That different thing might still be useful, it may even lead to the same or similar result, but it's deceptive to think of it as replacing the original action.