It looks like we are triggering the three second countdown on all SSL errors:

- Go to https://wrong.host.badssl.com/, it takes 3 seconds to see the interstitial.
- Reload the page. There is again a three second delay.

I'm seeing this on the corp network, so it's unlikely that it's waiting for a captive portal check. The only other code to trigger the delay is the network time tracker, and I'd expect it to return a response fairly quickly too.