When we use @RequestScope in a context where we would use @Stateless, we only simulate @Stateless behavior by destroying and creating the bean on every request. The benefit is we enforce isolation between requests; however, the tradeoff is we perform expensive creation / destroying of the bean every request.
If we care about performance and only need stateless behavior without enforcing isolation between requests, we should design our beans as stateless and use @Dependent or @ApplicationScoped scopes. There is no risk of having no isolation between requests if developers are careful and write their code statelessly.
Using @RequestScope as @Stateless sounds like a smelly workaround for a missing feature in the framework; hence I prefer not to use it in this context unless necessary.