If you have a lot of free clients or free client threads, then RR will disable these clients or threads.
The result is:
Jobs are send to the fastest online clients as first.
After all online clients are rendering, the 2nd job thread of a client is used, then the 3rd...
The rrClient checks the render log output while rendering for some special messages.
If it finds such an error message, it will abort the job.
E.g. Arnold plugin was not loaded (Makes no sense that Maya switches to the default renderer and continues)
The new setenv file for Maya copies the MtoA/Vray executable files to the local drive of the rrClient.
This way you have less traffic and less crashes if the network is busy.
The "wait for approval after preview" and "wait for approval after render" are now job based settings.
You can still set the default value in rrConfig for compatibility /upgrade reasons, but this default value can be overridden like any other job setting with submitter default files or render config files.
This is actually not a Royal Render feature, it is an NVidia feature.
NVidia has released new drivers in 2016 that you can use a CUDA 7.5+ renderer within a service. (Redshift 2.x works for example)
The minimum driver version for your graphics card has to be r361.
Note: We have not verified if it works with GeForce cards 7xx and older, but 9xx cards work fine.
A new rrCloudManager application is the connection between your RR installation and the Azure cloud.
It automatically creates/starts/stops your VMs and is able to create an OpenVPN tunnel.