We discuss recent progress concerning the resummation of large logarithms at next-to-leading power (NLP) in scattering processes near threshold. We begin by briefly reviewing the diagrammatic and SCET approach, which are used to derive factorization theorems for physical observables in this kinematic limit. Then, we focus on the quark-gluon channel in deep inelastic scattering and Drell-Yan. We show that the use of consistency conditions for the cancellation of leading poles in the hadronic cross section can be used to achieve the resummation of large leading logarithms (LLs) at NLP, both within diagrammatic and SCET methods. In this context it is also possible to investigate the problem of endpoint divergences appearing at NLP in SCET, and relate its solution to the concept of re-factorization.