In a 1922 paper, the intermediate of panel B is shown:
More than a 100 years later, this looks a bit funky.
A more recent paper from 1989 shows a dienone intermediate for the reaction in neutral or acidic aqueous solution. However, there is only one and not two bromine atoms in the para-position:
These are both screen-grabs from the freely available first page, thus the low quality.
It is possible that the mechanism varies with solvent and pH. Panel C shows the product of the first bromination step in its deprotonated form. Not sure why this would be called an intermediate, other than that the subsequent bromination steps are faster. This means that in the presence of an excess of bromine, you probably could not isolate the singly-substituted product.