When building a new feature for your website, avoid using the legacy platform that the rest of the site is built on.
Legacy integrations take longer. Your goal with a prototype is to validate if it's worthwhile. This will take more work if your integration exposes bugs or side effects in legacy systems.
New designs get coupled to legacy issues. For instance, you may have a complex registration funnel for your main product. But your new feature may not require that, and might be better off with a simple process. You may find that the simpler registration feeds the more complex multi stage process, rather than pushing full registration for a new feature.
Deep integrations can mask issues with your prototype. You may see user adoption, but only because the new feature is very visible to your installed base. The prototype might be flawed and produce false positive results because of the legacy integration.
Modern platforms with better features are ignored. You may have a poor search function on your site. Requiring the prototype to use the legacy search isn't really necessary. You can take advantage of newer versions of software by limiting their use to your prototype.
You can break legacy inertia with prototypes. If your prototype requires newer versions of platform components, you will build more company expertise in the new version. This will make it more likely that you can migrate legacy code to new versions.
Build new prototypes in isolation, so that you can develop faster, learn more and make better product decisions.