Ah I see. The equations I'm talking about:
[tex]\mathbf{F}^{ms}(\mathbf{x})=\frac{\mu_0 I I'}{4 \pi} \oint_{C} d\mathbf{l} \times \oint_{C'} \mathbf{l'} \times \frac{\mathbf{x}\mathbf{x}'}{\mathbf{x}\mathbf{x}'^3} [/tex](1.11)
[tex]d \mathbf{B}^{stat}(\mathbf{x}) \equiv \frac{\mathbf{F}^{ms}(\mathbf{x})}{I} =\frac{\mu_0}{4 \pi} d\mathbf{i}'(\mathbf{x'}) \times \frac{\mathbf{x}\mathbf{x}'}{\mathbf{x}\mathbf{x}'^3} [/tex] (1.15)
where [itex]d\mathbf{i}(\mathbf{x}')=I d\mathbf{l}' (\mathbf{x'})[/itex]
I'm afraid I still don't understand your answer to 1). Suppose "small" means "small compared to the characteristic scale over which the Bfield varies" (which Thide couldn't say at this point). That doesn't make a difference, because the tangent vector to the loop still turns through a complete revolution (!), so no matter how small it is, the corresponding force still pushes different elements in different directions, right?
