The web normally works by pulling assets: as soon as the browser knows that it will need a script, stylesheet or image, it pulls it. HTTP/2 Push allows the server to take the initiative by sending assets it knows the browser is going to need ahead of time. The challenge is making good use of HTTP/2 Push in the way ShimmerCat can do thanks to the AI engine that analyses data and create optimisation rules.
ShimmerCat uses HTTP/2 Push in conjunction with cache digests and asset prediction to automatically reduce loading times in a way that no other web server does.
The short answer is that it works well: if a resource is in the browser’s cache, there are ways to avoid pushing it. Our implementation of HTTP/2 Push is very good at avoiding reiterated push.
Below you can see the effect of HTTP/2 Push in a waterfall view: