There is certainly opportunity, and I should suggest up-front that some of this has shifted since last year's brainstorm on the same topic. For example, the related items widget is now much more usable, so some of the suggested improvements (or a suggested rewrite) from 2016 are not really that interesting from an impact/benefit standpoint.
My suggestion, in addition to the items @cewing recommended, is to experiment with patternslib alone in a plain-HTML (not Plone) context, and get familiar with both patternslib itself. Then get a sense for the functionality of the stock/demonstrated widgets on demo pages for each of patternslib and mockup. At this point, getting a git checkout of the mockup repository is likely a good idea, so you can start to understand the build process to create the (built) "bundles" that mockup creates.
Also, a tip for any prospective proposal: any realistic project is likely taking on 2-4 widgets or components, not every prospective improvement on a large list. The number of components that could realistically be done in a GSoC project will vary on their complexity. I'm particular to the items I enumerated here (on the first point of that list), but that's just my opinion, and may not reflect the pain points or needs of the community at large.
Sean