Finite-time degeneration for variants of Teichm\"uller harmonic map flow

We consider the question of whether solutions of variants of Teichm\"uller harmonic map flow from surfaces $M$ to general targets can degenerate in finite time. For the original flow from closed surfaces of genus at least $2$, as well as the flow from cylinders, we prove that such a finite-time degeneration must occur in situations where the image of thin collars is `stretching out' at a rate of at least $\text{inj}(M,g)^{-(\frac14+\delta)}$, and we construct targets in which the flow from cylinders must indeed degenerate in finite time. For the rescaled Teichm\"uller harmonic map flow, the condition that the image stretches out is not only sufficient but also necessary and we prove the following sharp result: Solutions of the rescaled flow cannot degenerate in finite time if the image stretches out at a rate of no more than $\lvert\log(\text{inj}(M,g))\rvert^{\frac12}$, but must degenerate in finite time if it stretches out at a rate of at least $\lvert\log(\text{inj}(M,g))\rvert^{\frac12+\delta}$ for some $\delta>0$.