Here is a quote from Neubert himself:
What I am going to propose (..) could be of some help both as insight and tool. Insight in the sense that an analysis of the textual structure of translated and/or interpreted stretches of discourse supplies information about text production and comprehesion in general. Tool in the sense that the treatment of the translation process and its results as intertextual strategy can serve to improve language mediation [i.e. translation, TH] itself and, above all (and that is, incidentally, my daily job), to make the teaching of translating and interpreting more effective. (Neubert 1984: 149)
Text linguistics can help us understand how texts tend to be put together in different genres and cultures. It can also show how those who are addressed by those texts comprehend and reproduce them. By studying the discourse of translators and interpreters, the researcher can arrive at an idea of how their cultures typically form texts belonging to a given genre. That can help to train translators.
A very simple example. In some traditions it is ‘not done’ to write academic essays in the first person singular (‘I think…’, ‘I want to argue…’). Instead, students use impersonal constructions, plurals or passives (‘One can say that…’, ‘We want to claim…’, ‘It will be argued that…’, etc.). In translating such an essay, the translator may automatically adjust the original. Text linguistics would begin by drawing attention to the different conventions and might then ask the trainee translator to think about what, in a specific situation, would be the most appropriate rendering.