Most of current ring-based Ethernet protection schemes work on a basis of filtering database (FDB) flush operation to guarantee data forwarding on protection switching. However, when a link or node failure occurs in a congested ring network, delayed data frames that have lower priorities than control frames can cause the FDB inconsistency problem that ring nodes build incorrect FDB information. Under this circumstance, some of ring nodes forward data frames in the wrong directions, so that the ring network shows undesirable and unstable performance. In order to resolve this problem, this paper introduces the following three techniques: flush delay timer, purge triggering, and priority setting. Especially, for the priority setting scheme we develop an absorbing Markov chain model for the Ethernet Ring Protection (ERP) protocol defined by G.8032. Then, we theoretically estimate the expected protection time on protection switching underlying the M/M/1/k queue system and also suggest an enhanced priority setting method to guarantee reliable protection switching even in a congested network. Along with our theoretical approach, we additionally evaluate the protection performance with a network simulator to verify our mathematical model. Simulation results show that our analytic model is indeed capable of estimating the expected protection switching time.