Markov-recapture models obtain estimates of insect population size by concurrent operation of traps self-marking bait stations, with capture probabilities obtained from a Markov-process model of movement into the trap. Estimates are readily improved by repeating observations over time. However, a sequence of observations may reveal violated assumptions in ways not possible for a single observation, and it may not be possible to discern those violations from a fit of data. This article uses simulation to examine the effect of the following 5 types of violations: (1) using the estimate in an open population, (2) individuals are not uniformly catchable, (3) marking and trapping rates vary over time, (4) unequal response to marking stations and traps, and (5) behavior changes after marking. Individual catchability and behavior changes caused small biases, and were generally missed by a test of fit. All others could cause large biases. Among these, variation over time was easily discliminated by a test of fit. An open population mayor may not be discerned by a test of fit, and should be accepted or rejected on the basis of independent data. Finally, the rate violation is rarely picked up by a test of fit. It is suggested that it may be possible to estimate the parameters for the rate violation from the experimental data, and adjust the population estimate and confidence limits accordingly.