"For a large one, 20% accuracy is science fiction stuff compared to the actual accuracy rates of big projects.." (exactly, I made that point as well) "..especially the kind with changing requirements, new issues discovered en route, etc (which is most of them)." If you are working in an environment where shipping something at some random point in the future is fine, well, good for you. The rest of us doesn't work in that world, which means that if you start a large project, lets say over the scale of 1 year, you cannot not talk about progress during that year. This is the point that I want to make.
In that sense: "5 hours spent on X" means, dishonesty aside, that 5 hours have been spent. This has some meaning, namely:
- a) if the estimate has been 1 hour, and this doesn't cancel out over time there is a systemic problem which puts the entire estimate to question. So someone should do something about that, and, whatever that something is - adding new or different resources, changing the goal, changing the timeline, whatever), it is usually not something that developers could do. (Not that I think 5 hours is a useful estimate size for any feature; 1 week, on the other hand, for a set of features would probably be useful)
- and b) if you figure out that team members are frequently spending more than 8 hours per day for whatever features (adjust 8 hours to whatever the team agrees upon) then the team is overloaded, the workload is not sustainable, and, again someone should do something about that...
Really, it shouldn't come as a surprise that - opposite to developers' lore - managers really do add value, especially when they are good managers.