As with any stall value, you'd want it as low as possible, and by that, I mean under 5ms would be idea in most cases, in some cases that would...

SQL STD AG is a basic availability group, with no read only routing. Still sounds like your application is connecting to a node name and not a listener name.

If the 2 instances are on the same machine, did you allow for the 2016 instance to have sufficient memory, assuming you allocated everything to your 2012 instance previously?
set...

Connect to your listener name and run SELECT @@SERVERNAME to confirm you are on your secondary node.

You'll need to make use of Powershell on the Windows Cluster level, as that is where preferred nodes are set, and build you alerting around it.
https://technet.microsoft.com/en-us/library/ee460989.aspx - this...

You need traceflags, 1117 and 1118 - read up on each first before implementing. 
1117 would keep your files growing equally - https://support.microsoft.com/en-us/help/2154845/recommendations-to-reduce-allocation-contention-in-sql-server-tempdb-d

Had to check, thanks for the question.

You will need to provide a lot more for anyone to assist you, well, with any quality replies. 
Post DDL, sample data, what you have tried so far.

That changes things, but you can do BAGs, Basic Availability Groups still, one DB per group, EDIT: in 2016 only again.
I know this might not be a solution for...

If you can afford to carve off more storage, my vote would go to Availability Groups instead on 2 x standalone VMs.
No shared storage required, just use a file...