Sudden problem with response times

So you have at least two remote repositories (the NAS with 4000 plots and a server with 12 full units [I assume you mean 12 full drives])

When you say textures, frames and units, it is like you are talking in code. What we need to know is how many plots you have an where they are.

If you have multiple remote repositories with that many plots being farmed by one node, that is a problem and will cause the issue you are having. See what worked for me in the post below.