Google indexer is two-pass, first run is without js

A comment from JohnMu is even more informative:

Yeah, there's no fixed timeframe -- the rendering can happen fairly quickly in some cases, but usually it's on the order of days to a few weeks even. If your site produces new / updated content frequently & you want it indexed quickly, you need that content in the HTML.

Does this really surprise anyone? Why should my scraper have to execute your arbitrary code bundle, and why should I as a reader have to run your bundle to get any content? The default dom with content should be the one that gets delivered.