The midpoint torsion bound #
The revised paper lemma lem-midpointTorsion excludes every positive even
period through 2g - 2, and hence every positive period below g, for a
length-two midpoint paired with a non-midpoint on another strand.
We use the paper's quotient/remainder reduction and Dhar argument, via the existing reduced banana normal forms. Reducing at either endpoint also handles the reflected case without choosing an orientation. In fact the even-period argument only needs the first mark to be a midpoint, of any strand length.
The even-period midpoint bound, including either side of the second strand's midpoint. The first marked strand need not have length two.
Doubling a putative period gives the second assertion of
lem-midpointTorsion, without an evenness assumption.
The length-two branch of Proposition 4.19, now deduced from the stronger midpoint lemma rather than initial-slope estimates.