If you want to only overwrite a single hosted feature layer, one option is to publish each feature class as separate hosted feature services, that way, if you overwrite the study area HFS, it won't clobber the wetland or streams feature layer(s) as they would be in thier own HFS.
Since there are issues beyond our control in AGOL, and, no real backups, any important data that I have in AGOL gets copied locally on a regular basis. For the most part, my script just backs up the HFS's to local FGDB's with date appened to the name. This way, I at least have a snapshot backup if 'something' happens.
For other layers that I tend to overwrite from Pro, my scheduled script checks for features that are in the AGOL hosted layer, but not in my local data (that I originally shared the layer from), and then appends the data to the local dataset. This way, if I overwrite the layer, it has ALL the data in it.
Kind of clunky method, but it gets the job done. Hopefully others will have 'better' options.
R_