This paper presents the design of a software-derived backward error recovery technique to overcome transient sequencing-errors in microprogrammed systems, with particular reference to the microsequencer of the AMD Am2900 bit-slice microprocessor family. Evaluations of the hardware, microstore, and execution time overhead requirements are given. The authors conclude that the overhead of hardware, storage, and time is reasonable when compared with that of full duplication. This is a practical, concise, and readable paper that is accessible to nonspecialists. The theoretician in fault-tolerant computing stands to gain practical insight into problems associated with reduction to practice from this paper.