Does the goldfish grow as big as the tank?

We all know about work extending past expected delivery times (to often the case), but how often are we confronted with early delivery? I mean early quality delivery? There's a saying - not sure from who - about the goldfish growing as large as the tank. When providing estimates, we all assume that there is some 'buffer' built in. How often is this buffer consumed by 'additional' work being provided, beyond the expectations? Over delivery - should this be considered the same type of issue as under delivery? In many aspects the answer is .... not sure. Over delivered product is often looked upon positively by all and sometimes nejavascript:void(0)eded to 'make up' for prior under deliveries. However, by over delivering on something, is there a risk that other critical tasks are not getting the attention that they should be. Is over delivery often on items that are non-critical? easy to do? clamorous and/or fun? Is this an indication that some core task, that is more difficult to delivery, is not getting the attention it should be?

I think the overall response to either over or under delivery is ensuring that a task's delivery expectations in regards to date, quality, results, etc. need to be clearly defined and if we (PM's) notice a consistent over delivery - that we need to review all work to make sure proper attention is given to all and that the over delivery is not a sign of other issues.