A bloated JPEG is definitely smaller than a TIFF, but 2x file size is hard to swallow for metadata, especially when most people aren’t that good about it. My most recent client project was all sourced from iPhoto and Photos libraries from multiple computers, consisting of 1.3TB of content. Doubling that file size just to get some keywords isn’t a reasonable tradeoff for me.
Instead, my approach is to extract content from the Apple sources and apply industry-standard metadata for what the client cares about. I can thoroughly tag 500-1000 photos an hour, and I can always open the Photos library for reference if I need some clues.
Your idea bout syncing through iCloud for Windows is super interesting. I wouldn’t want to sync 1.3TB of photos from a Mac, through the cloud, and to a Windows machine just to get keywords. But I will test and see if it works out of curiosity. As much as I don’t like Apple Photos, I think Windows is the real “turd soup” so I don’t go there if I don’t have to. 😉