Since I was participating in the original Q&A with some comments and the first to vote too, I'll add my two cents as well. 

I completely agree that the **duplicate question** is not quite the same as yours, but implied in your question is that you have performance problem with your rsync job, which you're trying to resolve. 

The fact is that rsync version 2 is still widely in use (and your question didn't mention the version of rsync you'd been using). Therefore **the duplicate answer** in the linked question explains quite well the performance improvements made in rsync version 3 and the conditions that need to apply to actually benefit from them (which are relevant even when you already run version 3). 

That for me was the reason to suggest that in fact it *might be* a duplicate Q&A.

Your comments clarified that a little but by then the [peer review](https://serverfault.com/help/privileges/close-questions) process can easily overlook those unless you edit/update your question. 

Even re-reading reading the question now it still gives *me* the impression that it potentially is an [XY problem](http://xyproblem.info/) where rather than solving the implied underlying performance problem you appeared to be asking about a non-standard work-around, effectively how to make rsync not act like rsync.   

Rsync is of the "trust but verify" school of thought and if you don't want/need that you can still somewhat use the rsync protocol, which was what I focused on, but the alternative is to simply not use rsync, but for instance use actual back-up software and do [incremental-forever backups](https://en.wikipedia.org/wiki/Incremental_backup#Incrementals_forever). 

And a final thought: a (single) replica is not often a substitute for a back-up.