robust in this context means that the theory is non-testable. It can adapt to cover any evidence or lack of evidence, hence it is a meaningless theory which cannot be falsified, cannot be tested and cannot predict anything.
It is testable, and not all science involves direct observation. It can be falsified. Say, if you find a human fossil in rock that is undoubtedly from 80 million years ago. The scientists would have no choice but to change the theory, or abandon it altogether. This is the nature of science. I'm sure there are scientists out there wishing they could find this, because they would be one of the the most famous scientists of the last 100 years.
Take Creation Science, however. It's non-testable unless God starts creating stuff again. It doesn't have to adapt to new evidence because new evidence must adapt to it. In opposition to scientific method, it cannot be falsified because the first tenet is that it is true (while evolution is that this is the best we know based on what knowledge we have, subject to change). And it can't predict beyond what's already written.
Seriously, evolution has been around a long time, and while small parts have been challenged and changed over the years, the basic concept hasn't been successfully challenged yet in the open peer-review scientific community.