@colintd I appreciate the follow-up, as I had been meaning to post back here as well.
Happily (for me), it's no longer getting in my way since I did ultimately decide to port my project to using STM32Cube HAL API instead of Arduino Wire APIs. After that, I've had zero such issues.
I realize that doesn't help folks much who are stuck using Arduino APIs for whatever reasons (or prefer those APIs -- there are some advantages).
Your theory is interesting, though. I honestly don't remember how deeply I looked into inter-call delays and/or deeper status checking at the time. I could believe that perhaps the Arduino Wire API implementations on different hardware could also lead to subtle timing differences... and if it is also specific to the M5 PPS device, those docs aren't ultra-specific about what the device actually expects w.r.t. delays, etc.
TL;DR: It was an interesting side quest, which I ultimately solved by changing middlewares. ╰(°▽°)╯