It's real. We've landed enormous performance improvements this year, including migrating most Firefox users to a full multi-process architecture, as well as integrating parts of the Servo parallel browser engine project into Firefox. There are still many improvements yet-to-land, but in most cases we're on track for Firefox 57 in November.
I hope not. We've been working our butts off optimizing stuff across the board. The big projects (Stylo, WebRender) are ongoing and awesome, but there's also a legion of people working on eliminating boring papercut performance issues (a few milliseconds here, a few milliseconds there).
Still have more stuff to land, so expect things to keep getting better.