Some protocols are sensitive to mere collisions. Also, this suggests actual preimage attacks could be found later. Not today, but still. Finally an accurate explanation is longer and has less punch.
It's safer to "join the headless chicken" route, consider SHA-1 officially broken, and start thinking of alternatives.
IIRC, collision attacks may be considered precursors to preimage attacks. Some preimage attacks look like "brute force" monte carlo simulations of collision attacks, running enough collision attacks in parallel until a collision attack results in the preimage.
It's safer to "join the headless chicken" route, consider SHA-1 officially broken, and start thinking of alternatives.