When R1 restarts after a failure, it does not route traffic for R2, nor does it provide backup for R2, until the hold-down timer expires. Similar to VRRP, the hold-down timer value must be greater than the time the routing protocol requires to converge its tables.
During the hold-down interval, R1 routes traffic if the destination MAC of the packet is its own routable VLAN MAC. R1 bridges incoming traffic to R2 if the destination MAC of the packet is the routable VLAN MAC of R2. A temporary default route (one having a route preference equal to 4) that points to R2 is installed on R1. R1 uses this temporary route to forward traffic to R2 that it cannot route itself because of the incomplete routing table; the default route is not saved in the configuration file.
After the hold-down timer expires, the temporary default route that points to R2 is deleted; from this moment on, in addition to routing packets destined to itself, R1 starts routing packets for R2, as well.