I'm still not 100% sure whether to suggest:

• "Dismantling" the Userbar component's template_name and get_context_data() to be used in the view, and let the view do the template rendering.
  • Pros:
    • the view is straightforward
    • the response uses Django's TemplateResponse, which means you can test it with things like self.assertTemplateUsed
  • Cons:
    • we're not really treating the userbar as a proper "template component that knows how to render itself"
    • If you subclass Userbar for some reason, the request can only be accessed in the parent_context of get_context_data(parent_context)
• Letting the Userbar do the rendering (i.e. via Userbar(...).render_html()) and make the view just use the rendered string as-is.
  • Pros:
    • we are treating the userbar as a proper "template component that knows how to render itself"
    • If you subclass Userbar for some reason, the request can be accessed after initialisation
  • Cons:
    • the view is less idiomatic Django (it needs to just return a response using a pre-rendered template string)
    • needs further refactoring of the Userbar class so it accepts request during init rather than inheriting it from the parent context in get_context_data

When writing this docs, I had the idea that we should probably do the second option. Though after looking into it, while I prefer having the request more explicit via init rather than inherited from the parent context, I wasn't convinced by the part where we need to "construct a Django response with a prerendered string" for the view. I think we should just stick with the current (former) approach.