There's not much we can do here separately, folding this bug into  issue 669192  "Measure relevant performance metrics".

IIRC, background was two separate issues:

- ensure there's essentially no performance impact when not showing any overlays or warnings on top of WebVR content. We had occasional regressions due to accidentally doing rendering steps for content that was supposed to be hidden, i.e. the loading indicator, but we can address these as they come up.

- when showing overlays, i.e. the http "not secure" warning, some performance impact is expected and seems unavoidable due to the extra composition needed. Ideally, we should use automated testing to ensure this remains reasonable.