I feel like this is the crux of the issue. Dev estimates should be taken with large error bars, and I think the entire problem is caused by people who take them at face value, making developers not want to give estimates at all next time.
Well, it helps if you have retrospectives, though I see them rarely done properly.
What kind of works is tracking similar efforts (in the same code base) and measuring how long they take. While every task is different, there are usually a small number of different classes of tasks, and knowing how quickly these can be done helps in both estimating as well as defending the estimates.
