In cases where a lot of people are contributing DNA to a sample, which I'd expect for most environmental sampling, it seems like you would need impractically large amounts of sequencing before you could identify anyone specific in sequencing conducted for other purposes: you'd need to get a sequencing read which could have come from only one person, or enough reads each matching a small number of people that you could make an inference about an individual.
The paper does have "The long eDNA reads sequenced here suggest that with targeted enrichment of informative genomic locations, one could achieve individual-level identification even from pooled samples" but this is just saying that environmental sampling with the specific goal of identifying individuals could be possible. This is very different from this being something that normal researchers might do by accident in the process of environmental sequencing for a different purpose.