Best reason for stub zones: they don't require maintenance. Let's say that your partner company makes a change to their DNS (ex. change of IP address), and, as is often the case, forgets to inform you about it. What happens? Well, if you're using conditional forwarding to direct requests to the partner's domain, those requests are going to start failing and you're not going to know why. If, instead, though, you are using stub zones instead of conditional forwarding, the new DNS info will be transferred automatically from the partner to you, resulting in no downtime; you wouldn't even realize that the partner made any change. That's one of the best cases of where stub zones are a better choice than conditional forwarding.

Another two reasons for stub zones: optimization and low processor utilization. Lists of forwarders take time to read and can be, consequently, processor-intensive, especially for long lists. Stub zones don't suffer from this.

So, if stub zones do what conditional forwarding does, only better, then when would you ever use conditional forwarding? It's hard to get a straight answer on this, but I think that it boils down to simplicity: forwarding is slightly easier to set up and may be more practical in some situations. That's my theory, anyways. What the experts seem to agree on, though, is that stub zones are almost always the better option and conditional forwarding should be used only when it's the most practical. I wouldn't worry much, however, since Microsoft seems to be more keen on you knowing stub zones vs. secondary zones and conditional forwarding vs. normal forwarding, rather than how the two new features stack up against each other.