I have the chronograph classic with the 89361 calibre, but the problems are similar (maybe we should start a thread on this base of calibres?). I use the chronograph every day and about one time in fifty or so actuations, the time stops. I don't use the flyback function, but presumably the same thing would happen if I did. I'll explain below why this doesn't bother me so much.
First off, apart from this peccadillo, this is a fantastic chronograph with extremely tight tolerances (in fact, I suspect this is the origin of the problem above). Of course, it's a column wheel, so it shouldn't be surprising that the timekeeping hands don't jump when the chronograph is actuated. What is surprising, however, is that it's the only watch I own where the hands don't jump when the crown is pushed in (I like to have the second hands perfectly aligned with the minute hands, so this is a big deal for me).
Everything just feels very solid. Moreover, the hour and minute hands on the chronograph register are very well aligned (not perfectly, mind you), much more so than any of the Valjoux 7750 based chronographs, and even better than my 4130 calibre Daytona which for me, is the standard bearer. (Given this last fact, I'm surprised about additional complaints regarding hands not returning to zero on the chronograph register. Presumably, if it's consistently returning to the same point, it's an issue with the hands and not the movement.)
All this suggests to me that the gear tolerances are very small on this watch. Most watches have some wiggle room; presumably this is a design feature to allow for smooth operation. My 89361, on the other hand, runs like it's on rails. This is a great thing, but I wonder if it leads to the stopping problem that seems to be widespread; i.e., the watch operates on a knife edge and the chronograph actuation is a sufficient perturbation to stop it. I suspect this is a near universal problem, so won't return my watch, and instead just enjoy the up side to having such a "tightly wound" calibre. I hope that IWC can fix this in future iterations, though.