I believe that the most common cause of this is that you have a loop in your flow somewhere so that the nodes never get the processor to action messages. If you look at the processor utilisation do you see that it is heavily utilised when node-red is running? If so then possibly you have an MQTT loop so that you are publishing to a topic, subscribing to that topic and the value received causes another publish etc, etc.
If you look at, most of the cases are going directly from a MQTT message o a smooth (with the error) and a graph, so there is nothing in there than can make a loop.
If you take disconnect the inputs to the the smooth nodes and feed the unsmoothed values to the dashboard instead, does it all work as expected? If so then try adding new smooth nodes configured the same as the old ones.
It will take some time (maybe days) to get it done and see if problem comes again after receiving a lot of data but I will try and will return the feedback.
Thanks a lot for your time Colin, much apreciated.