Long-tract gene conversions (LTGC) can result from the repair of collapsed replication forks, and several mechanisms have been proposed to explain how the repair process produces this outcome. We studied LTGC events produced from repair collapsed forks at yeast fragile site FS2. Our analysis included chromosome sizing by contour-clamped homogeneous electric field electrophoresis, next-generation whole-genome sequencing, and Sanger sequencing across repair event junctions. We compared the sequence and structure of LTGC events in our cells to the expected qualities of LTGC events generated by proposed mechanisms. Our evidence indicates that some LTGC events arise from half-crossover during BIR, some LTGC events arise from gap repair, and some LTGC events can be explained by either gap repair or “late” template switch during BIR. Also based on our data, we propose that models of collapsed replication forks be revised to show not a one-end double-strand break (DSB), but rather a two-end DSB in which the ends are separated in time and subject to gap repair.