DNA double-strand breaks (DSBs) are potentially lethal lesions that arise spontaneously during normal cellular metabolism, as a consequence of environmental genotoxins or radiation, or during programmed recombination processes. Repair of DSBs by homologous recombination generally occurs by gene conversion resulting from transfer of information from an intact donor duplex to both ends of the break site of the broken chromosome. In mitotic cells, gene conversion is rarely associated with reciprocal exchange and thus limits loss of heterozygosity for markers downstream of the site of repair and restricts potentially deleterious chromosome rearrangements. DSBs that arise by replication fork collapse or by erosion of uncapped telomeres have only one free end and are thought to repair by strand invasion into a homologous duplex DNA followed by replication to the chromosome end (break-induced replication, BIR). BIR from one of the two ends of a DSB would result in loss of heterozygosity, suggesting that BIR is suppressed when DSBs have two ends so that repair occurs by the more conservative gene conversion mechanism. Here we show that BIR can occur by several rounds of strand invasion, DNA synthesis and dissociation. We further show that chromosome rearrangements can occur during BIR if dissociation and reinvasion occur within dispersed repeated sequences. This dynamic process could function to promote gene conversion by capture of the displaced invading strand at two-ended DSBs to prevent BIR.