We ended up reworking the formatting of our HTML-to-PDF export. We were using the former default charset=windows-1252 (as seen in the screenshot):

We updated it to UTF-8 and tightened up the HTML formatting and made a few other adjustments. The output looks right again! Took me and a developer a few hours to get it sorted out once we identified the charset as the most likely culprit. 
So it seems like Microsoft enforced some updated standards for HTML encoding in the Power Apps ecosystem. This probably affects people who've used Word or Outlook to create rich text to begin with.
Ironically, Microsoft's HTML code, when created from these apps, is horrific and poorly formatted. Best practice is to cleanse it of spans and in-line formatting first, apply CSS, and then use that HTML in your template. Or just write it with Copilot.