Performance wise, there aren't many bottlenecks right now. Switching in Resque for my job queue (replacing Delayed Job) will make everything a little faster. If the queue ever starts to get large, I can quickly fire up another server as a dedicated worker.
I am leveraging AWS Cloudfront as a CDN for the widget JS, so the widget will load super fast for everyone. I offload all static assets when I can. Google CDN for JQuery.
The only immediate point of concern is the real-time API. That will soon have some throttling/limits on it to prevent abuse. I highly prefer people use the Post-back method to get notified of verifications (push is always better than poll).
As for proving reliability, I'm thinking of putting up public Pingdom reports for transparency.