The one you uploaded seems like it's a bit blurry, this one is pretty sharp.
He probably 'Open image in new tab'ed instead of pressing the blue button which loads the original size. Sometimes Fantia upscales the previews to fit the screen.
So from this point onwards it should be considered superfluous and not worth uploading pixiv images without a significant size upgrade, if not outright spam?
If he doesn't change how he uploads to pixiv, then yes I'd consider it superfluous.
I'd be very interested in knowing how gao managed to upload a jpg losslessly to twitter. There doesn't seem to be any sort of global changes to how compression works, looking at other recent twitter images... Some juz images also recently went up on twitter without the horrid compression that they normally have.
I changed the parenting because the images seem to be straight re-uploads of the twitter image to pixiv by the artist, with the slight difference in file size because of pixiv re-encoding the image.
Thus, in terms of both compression guidelines and first-to-the-post relations, I'm sure the twitter images should be the parents.
Edit: That is to say, the artist is not uploading the source image separately to twitter/pixiv, he's uploading the twitter image itself to pixiv.
@lxs What I've found is that plurk images have more jpg artifacting but has (much) better red compression than twitter. It's why a bunch of artists prefer to upload pngs or host imgur links than upload jpgs to plurk. I say if an image has enough red, the plurk is preferred.