...no way of telling from the system side...
A server can only tell the difference if the client cooperates. That is, there is nothing in the connection itself that indicates how the IP address was obtained.
The connection is always by IP address. By the time a connection is requested, any work that was done by the client to obtain a resolved address is long since been done.
You could run programming on the server that called back to the client and requested data about how the address was obtained. The client would, of course, need complementary programming that could respond.
Other than something like that, the only way I can think of is like I said before -- enable a second IP address on the server and point DNS at the new address. Clients that continued to connect to the old address are the ones not resolving through DNS. If all clients automatically use the new address, then you don't need to fix any of them.
You Describe my issue perfectly in your latest answer, I have a Prod and a HA system and I wish to role swap between the two. The issue I have is that I need to identify those users who have had to change their desk top icons over the years for whatever reason to hard code the IP’s.
During the Roleswap part of the changes we make is to make the DNS changes to point the users seamlessly to the other server. Obviously the users with Hardcoded IP’s will have issues. I want to identify these users now for two reasons,
1. they need changing, but
2. to understand why they are IP hardcoded, they may have an issue with the DNS resolution which I need to fix before I can role swap.
While you sound like you know what I need to do, I have to apologise as I am not sure I understand the resolution, if indeed this is a resolution for this scenario.
I should also mention that my HA environments are locked down when not being used in Production, we have a separate QINTER for my support team so we take down QINTER itself and also close down things like TCP/IP interfaces etc. We do this to stop our clever “IT” users access the systems and messing up my replicated data sets.