Basically semantics about where the "winding" is. Note that he indicates
0.9V - (-0.1V) = 1V for the total loop voltage; Kirchoff isn't violated if
the EMF is placed symbolically somewhere in the loop. Where doesn't matter,
and in reality, it will be evenly distributed (of course, we don't have
non-Laplacian fields, either, so there will be a nonzero amount in the
voltmeter loops as well).
Calling it "nonintuitive" is silly. There is nothing nonintuitive about
Maxwell's equations (at least until you start throwing nonlinear materials
around). E&M is *easy*.
The people I really feel bad for are the people involved in any fluids:
aerodynamics, fluid mechanics, hydraulics, etc. The Navier-Stokes equations
can be linearly approximated in only a limited set of cases -- small
Reynolds number. Any real, interesting and useful application is hopelessly
difficult to analyze, and, until recently, was mostly confined to empirical
data (this flow through this roughness of pipe requires so-and-so pressure
drop) and trial-and-error experiments. They tried, what, hundreds of
combustion chamber designs on the Saturn V rocket before they found one that
didn't KATO? Nowadays, these systems can be simulated, at huge
computational expense, still involving many turbulence and scale-related
approximations to try to optimize the process, but the development of models
is still largely dependent on the user.
Tim