For a few decades after molecular manufacturing was first described in detail, a variety of scientists looked for, and thought they found, a variety of reasons why it couldn't work.
In the end, it turned out that math trumps intuition - and the math supports molecular manufacturing.
For example, scientists who studied unattached atoms on surfaces knew that those atoms sometimes move around at room temperature. But the fact that an unattached atom will move says almost nothing about whether a firmly bonded atom will move - in fact, it's pretty easy to calculate that in most cases, it won't move. Atoms are slippery, but not that slippery.
Of course, even doing math can sometimes get you into trouble, if you don't check your assumptions. Some scientists dug up a theorem that appeared to prove that a flat graphene sheet could not exist, unless it was supported by something. (Graphene is just a single layer of graphite.) In other words, if you stretched a graphene sheet across a hole, something would break it or bend it or otherwise make it unstable. Since graphene has been proposed as a construction material for nanomachines, this might be a problem.
But in fact, the theorem doesn't say what the scientists thought. All it says is that the spacing between the carbon atoms in the graphite will not be quite regular over long distances. In effect, the sheet will be distorted sideways, randomly, a little bit, in patches. This is a pretty boring result, for most engineering purposes. As far as I know, it does not affect any proposed nanodevices.
For a more technical explanation, see Drexler's explanation of the issue.