By the way, in almost 100% of cases (the rest being just OA where the published version could be sent by anyone to anyone or something legally really dubious), the authors have a right to send their paper, even if it is published in a paywalled journal. Basically, the only thing the journal has a right to for subscription-based (aka those that cost $35) articles is content plus page layout. If the authors have the exact same text but formatted differently, they are free to distribute it wherever and however they want.
Preprint servers or lab/personal websites are best first choices for that.
edit: a small disclaimer on the exact same text meaning exact same text the authors provided; if the editor in the journal has corrected some typos and inserted a/the here or there (a common thing for non-natives to miss), then this becomes more of a grey area, because technically at this point it’s not a 100% authors’ text).
I’d say (a couple years ago) the service is also supposed to be access via DOI in perpetuity and presence in all the relevant databases, so that’s gotta cost some money for the reassurance as opposed to a pdf file “hosted” on Google Drive. But after Heterocycles fiasco I am not sure about that anymore.
Well, and some mark that this is likely a valid piece of research if it’s at www.reputablejournal.com as opposed to this likely being half-baked something at www.somerxiv.com or this likely being absolute lunacy at www.anyothersite.com.
Still, yes, billions in revenue vs millions spent essentially on essentially simple tasks like hosting and cataloguing (plus matching authors to reviewers I guess, though with how often I am asked to find them myself it’s doubtful) does not compute indeed.