This seems all quite speculative. In the first paragraph they describe FLoc IDs as changing constantly - why would they assume new IDs are not being generated, and groups are not constantly being mixed?
> “If your behavior doesn’t change, the algorithm will keep assigning you in that same cohort, so some users will have a persistent FLoC ID associated with them — or could."
When combined with other information that is already being used (such as canvas fingerprinting and other techniques), this looks like it can help narrow it down even further.
> “We can use that as another signal to create a stable identifier for them.”
The best way to explain it is that a lot of companies have been making significantly more money than their technology is worth. A number of initiatives have attacked the data-in side of the equation so the underlying tech is showing how questionable it really is.
This type of research should be filed under “could be big” but at this point it’s closer to public relations for ad tech firms than “the sky is falling, become Amish.”