Hi there,
ONTAP maintains a per-client set of synchronised locks for enabling continuously available shares, which use up NVRAM on each controller in a HA pair serving data.
With a limited number of HyperV servers using a limited number of files, we've been able to model and test it to our satisfaction.
With several thousand general purpose clients, opening a totally unknown set of files, it's more difficult to model, and situations may arise where the feature slows down the controllers too much (by eating NVRAM and forcing more frequent checkpoints).
For "general purpose" use, the clients will generally re-try the connection and deal with the slight interruption better than an enterprise app running inside a Hyper-V or SQL environment would.
We know we have customers that have turned on CA for general purpose storage, and don't have problems, sometimes in very large environments, but as we tend to be conservative (as most storage admins are), it is not a "supported" use case. We have fixed BURTs that only occur with CA shares being used outside this use-case, so if it can be proved that there is a technical issue (vs performance) only present when it is enabled, you may raise a support case.
I hope this helps with your analysis of if this feature is right for your environment. Please let me know if you have any further questions.