Small-n fMRI studies tend to have issues with replication because of how noisy and difficult to use fMRI machines are. I would take this research more as an indication that more research is needed, as opposed to the be-all and end-all of how short-form video interacts with the brain.

(Caveat: I have no expertise with neuroscience or using fMRI machines. I'm mostly basing my comment off pre-replication crisis fMRI experiments that failed to produce useful data upon scrutiny.)