That's what I tend to recommend. Easier and more consistent than debugging the semi-regular failure of repl after a deployment.