The trade-off hypothesis proposes that the evolution of pathogens' virulence is shaped by a link between virulence and contagiousness. This link is often assumed to come from the fact that pathogens are contagious only if they can reach high parasitic load in the infected host. In this paper we present an experimental test of the hypothesis that selection on fast replication can affect virulence. In a serial passage experiment, we selected 80 lines of the bacterial insect-pathogen Xenorhabdus nematophila to multiply fast in an artificial culture medium. This selection resulted in shortened lag phase in our selected bacteria. We then injected these bacteria into insects and observed an increase in virulence. This could be taken as a sign that virulence in Xenorhabdus is linked to fast multiplication. But we found, among the selected lineages, either no link or a positive correlation between lag duration and virulence: the most virulent bacteria were the last to start multiplying. We then surveyed phenotypes that are under the control of the flhDC super regulon, which has been shown to be involved in Xenorhabdus virulence. We found that, in one treatment, the flhDC regulon has evolved rapidly, but that the changes we observed were not connected to virulence. All together, these results indicate that virulence is, in Xenorhabdus as in many other pathogens, a multifactorial trait. Being able to grow fast is one way to be virulent. But other ways exist which renders the evolution of virulence hard to predict