Use two separate servers (VPS or whatever) as similar as you can make them (hardware and software) at the base image. Automate all configuration of your production environment so nothing is done by hand. Use that automation to produce a staging server that's identical to your production environment. Maintain the automation to ensure both environments stay in sync and can be replicated on demand.
Solves both your staging-out-of-sync problem and your first-order scaling problem.
As far as cost goes, VPSes are cheap as chips. The number of production downtime-inducing failures you'll avoid by having a staging server will pay for your staging environment in no time (unless you're not actually making any money at all, in which case downtime isn't so much of a problem and you can go nuts with the breakage).
