Error detection/correction techniques have been advocated for algorithmic self-assembly. Under rectilinear growth, it requires only two additional tiles, generally referred to as Isolation tiles. This process can be effectively utilized for checkpointing and is analyzed in this paper self-assembly. Initially, the physical framework (and related features) for the removal of the erroneous sections of an assembly is outlined. A novel Markov based model is presented to establish the optimal rate of checkpointing and assess its performance versus other error tolerant techniques that utilize redundancy. Simulation results are provided.