I haver not read the paper, but I assume they probed for frame-shifted products based on the slippery sites; triplet of 3 FU's (fake uridines)? Based on past work with nonsense suppression, we got read-though at stop codons using genuine, "all natural" uridine (perhaps feedback between translation and the nascent folding state). Wouldn't be surprised if paused-induced frame-shifting also occurs at single or double FU locales. Lot's of possible combinations, especially if you have frameshift followed by downstream frameshift. Bottom line is that you would need to screen for a lot more possible protein products to get a handle on frequency of the problem.