Distributed rendering requires some additional setup to be accomplished successfully. Rules on preparing the jobs change on a software basis and it also depends on the production workflow. By the way, we can provide some general rules to follow, if applicable:
Check files external references: Jobs often requires external files. When you pick up those files during the contents generation, you often pick them up from your local file system. In opposition, a distributed rendering requires contents to be available on the network. You should always pick up the external references from a network path, or consolidate them with your project structure, and then move the entire structure on the network
You should also consider baking data performed randomly like Global illumination photons, dynamic caches, particles caching and simila. By linking a cache object, you’ll be sure that any host will perform a valid frame that matches with the frames rendered across the farm. Also check the product documentation, there’re certain technologies that cannot be rendered on different hosts
If you experience strange results or general problems when rendering through Muster, we suggest you to test the batch render directly from the command line. Keep in mind that batch renders sometimes may have different behaviours than a rendering from the main software interface; you should always refer to the software documentation and check for differences/hints about using the batch rendering