Rules of thumb for capacitors always get engineers into trouble. Each dielectric system makes them very different devices and even within dielectric systems there is huge variance in how they perform.
polymorph:
If you start with a much higher voltage rating, the oxide layer may end up uneven
Actually, the oxide is already quite uneven. In order to increase the surface area of the dielectric, we acid etch the foil sheets to create valleys. Within those valleys, the oxide will always grow an uneven amount--either when we form it or when it is reformed during application use. It turns out that de-rating past 30% of the rated voltage generally doesn't increase the life of the part because of these areas constantly re-growing due to the localized heating that causes them to break down. (Just as you said, except that we already know it happens anyway.)
polymorph:
It is bad engineering practice to select a wet electrolyte aluminum or tantalum electrolytic capacitor with a voltage rating much higher
Again, you cannot apply the same reasoning to a wet-aluminum and a solid-tantalum, or even a solid polymer aluminum. For example, solids like mno2-ta, polymer-ta, and polymer-al don't have a situation where the oxide is breaking down over time.
I didn't mean for my argument to be that you you should use a 100V al-lytic in a 15V application. I was making the more general statement "use the highest rated voltage you can for your application." Which might mean considering many of the points you've brought up. Along with the practicality of most applications limits that rated voltage anyway.
polymorph:
If it were true that using a much higher voltage rating with a wet electrolyte meant longer life, then why do unused old capacitors require reforming?
You're mixing use conditions with shelf-life, which is important distinction. While sitting on a shelf unpowered, the oxide is being broken down by the electrolyte. Until voltage is applied and current can flow, the oxide can't regrow itself. So no matter what the voltage the dielectric was formed with, eventually the oxide will breakdown completely allowing enough leakage current to cause catastrophic failure when powered again.
So your example of 0V doesn't really apply because that isn't allowing the oxide to regrow.
polymorph:
http://www.eetimes.com/document.asp?doc_id=1279791
"Your first thought might be to increase the capacitor’s voltage rating to minimize the possibility of a dielectric failure. However, doing so can lead to a capacitor with a higher equivalent series resistance (ESR).
Yes, I agree with Rob's statements. However keep in mind that he doesn't not work for a capacitor company, and I do.