The WordCloudModule is rarely used, but accounts for about 160K worth of uncompressed JS in the 481KB XModule JS bundle. Pulling those JS files out of the bundle and serving them in the XBlock style should reduce the overall weight for the majority of students.
I was also thinking about what it would take to remove RequireJS Optimizer, and just have the files loaded individually. I see that it isn't just as straightforward as loading the files directly as there are other costs: https://blog.newrelic.com/2016/02/09/http2-best-practices-web-performance/. We should start doing the research as to what it would take.
FYI, word cloud is actually used quite a bit (shocking, I know).
Yeah, I was really surprised that it was being used at all! I thought it was a super-niche thing. But NR Insights reports that ~2% of all sequences loaded in the courseware in the past week on edx.org do have a word cloud module in them. Still, we should be able to see some worthwhile gains from splitting it off.
If anyone's interested in the NRQL for that (and comparison to Capa problems and Video):
SELECT filter(count, where `seq.block_counts.word_cloud` >= 1) / count, filter(count, where `seq.block_counts.problem` >= 1) / count, filter(count, where `seq.block_counts.video` >= 1) / count from Transaction where appName='prod-edx-edxapp-lms' and name='WebTransaction/Function/courseware.views.index:CoursewareIndex.get' SINCE 1 week ago
JIRA markup makes me want to scream.