If we're going to weigh the pros and cons of mesh over sculpties, let's look at it from three different perspectives: prim cost, upload cost, and the incompatibility of mesh imports with the viewers most SL users are able to use on their computers.
1.) Prim Cost
A single sculpted prim is just that: a single prim. Yes, the loading time for sculpties is atrocious and a region or parcel with many sculpted prims can seriously lag out a computer, often to the point of crashing. But in regions and parcels with limited prim allotment, the need to keep prim count down to within tolerable levels often outweighs whatever perceived benefits there might be to mesh imports, which have a higher prim count than normal sculpted prims. If an avatar full of mesh imports cannot be sustained within a region, what then? How quickly will parcels and regions max out their prim counts? The answer depends on how creative builders are, just as with sculpties.
2.) Upload Cost
As Moo pointed out, there is a higher upload cost for mesh imports than there is for a single sculpt map texture. For people importing a handful of mesh builds, especially ones with low(er) prim costs, this may be fine. But most people probably will not be able to afford the mesh imports required to heavily or completely revamp their lands. If you have surplus disposable income to spend on mesh, that is great. If not, you probably will find mesh prohibitively expensive.
3.) Viewer Incompatibility
This is probably the biggest issue with regard to mesh. With a majority of SL users finding Viewer 2 and its clones unuseable, they must find ways to stick with their Viewer 1-based viewers or leave SL altogether. This is not a very pleasing set of options to choose from, to say the least. Linden Lab has long sought to impose its inferior, outsourced viewer on the rest of the grid. Most Viewer 1-based viewers, which have consistently outcompeted Viewer 2 since the latter was created, cannot display mesh. This means that those of us who either cannot use Viewer 2 or simply prefer not to will not be able to see all the pretty mesh imports people are uploading unless they switch to a viewer that either doesn't work on their systems or which they can't stand on account of the poor design and exceedingly difficult user interface. To be sure, some third party viewer developers, such as Singularity, are working to incorporate features such as mesh into their projects. (As they succeed, count on more people to dump Viewer 2 and its clones.) In the meantime, however, most of the grid cannot see mesh, making it pretty much irrelevant.
If Linden Lab wants to retain customers, it needs to revamp its viewer so that it has the Viewer 1 layout and interface. This is not an impossible task, as others are successfully incorporating Viewer 2 code and features into their projects. It also needs to lower the costs associated with mesh in order for the investment put into it to see any significant kind of practical return on that investment.
Just my two cents, for what they're worth.