MR-based image-guided (IG) radiotherapy via all-in-one MR treatment units (MR-linacs) is one of the hottest topics in contemporary radiotherapy research. From ingenious engineering solutions to complex physical problems, researchers have developed machines with the promise of superior image quality, and all the advantages this may confer. Benefits include better tumour visualisation, online adaptation and the potential for image biomarker-based personalised RT. However, it is important to remember that the technical challenges are real. In many instances, they are skillfully managed rather than abolished, a point illustrated by the wide variety of MR-linac designs. The proposed benefits also deserve careful inspection. Better visibility of the primary tumour on an IG scan cannot be bad, but does not automatically equate to better IG, which often depends on a more generalised match to daily anatomy. MR-linac will undoubtedly be a rich milieu to search for IMBs, but these will need to be carefully validated, and similar work with CT-based biomarkers using existing, cheaper, and more widely available hardware is currently ongoing. Online adaptation is an attractive concept, but practicalities are complex, and more work is required to understand which patients will benefit from plan adaptation, and when. Finally, the issue of cost cannot be overlooked, nor can the research community's responsibilities to global healthcare inequalities. MR-linac is an exciting and ingenious technology, which merits both investment and research. It may not, however, have the future to itself.