I have recently managed to figure out a problem that has been bothering me for ages. Every once in a while, my mental ray renders would crash with a slave time out error when an image is distributed across my network using satellite. The problem turned out to be a very easy fix, and it seems to stem from the fact that when buckets are rendered too fast, the network or host computer gets overloaded by all the simultaneous requests or updates. All I had to do is make sure to add the -ts option to my render command, with larger bucket sizes that scale up with the size and simplicity of the rendered image. Thus, for masks, shadow, or depth passes that render extremely fast, I found that they typically require larger bucket sizes than the beauty pass for the same image. In these cases, I ended up having to go up to a 256 bucket size for images around 10k pixels large, while the beauty pass for the same image rendered fine at a 128 bucket size.