We worked a lot with service workers and it is rather meant as an improvement of an already optimized stack than a solution to solve the problems of the stack itself. Integrating a Service Worker in Volto is a no-brainer but we did not see the necessity yet. You have to be clear what you want to accomplish and how. Just implementing a Service Worker for Plone is not sufficient as a GSoC description IMHO.
The standard Plone frontend has more urgent issues in my opinion. I'd rather focus on @datakurre's Webpack approach first. Not sure if he is available as a mentor though. Once we are in the Webpack land with standard Plone, we can focus on improvements like service workers.
I agree with MrTango that we need more details what we want to accomplish with this.