Traffic to a RIP branch detours through the core after a reload

A legacy branch router, R1, runs RIPv2 only. Its LAN 172.16.10.0/24 hosts the ERP server 172.16.10.20. R2 and R3 each have a direct link to R1 running RIP, and an uplink to CORE1 in OSPF area 0. Both redistribute RIP into OSPF and OSPF into RIP (redistribute ospf 1 metric 3), with no route-maps.

R2 was reloaded for an IOS upgrade last night. This morning, users on R2's LAN (10.20.0.0/24) say the ERP system is slow. Last month, after a power cut at R3's site, it was R3's users who complained for a while. R2's direct link to R1 is up and clean.