When we are talking about predictivce/preventive maintenance etc with large amout of data, what you usually do is to setup an edge device that extracts the relevant data and does the analysis/calculations and send calculated data to an higher level system to act upon. All data is stored locally on the edge device and you push chunks of the data at regular intervals for further analysis. This way you can control when to load the network. In one of the projects i read about they had a network load scanner that constantly monitored the network and sent the saved database data when the network load was low it started to send large amount of data and if the network load started to increase it stopped sending.. Constantly pushing thousands of tags that you only need for "offline" evaluation/analysis is a bad idea. Push it at intervals or just leave them on the edge device.