Many consider the "ground" of the building to be impeccable "earth ground". I have only found one building with a perfect earth ground. That plant used a 300 foot well with all metal piping, and a meticulously bonded water sprinkler system. They used this system to ground all their systems to. This was the only plant that I never had any noise problems with.
Back then, I was supporting a PLC control system with PC and instrumentation. Looking back, I think the system design was inherently susceptible to noise. The design only worked well with a perfect ground. One of those hypothetical systems from the Deutschland.
So I am always skeptical when there is some odd intermittent behavior with controls and instrumentation. Throw in a few variable frequency drives (VFD) and some big motors and you add to the fun. Now add just a little corrosion or a few bad connections on the ground lugs and it gets real fun. Also keep in mind, these plants I am referring to can be 40 acres under one roof. One textile plant I installed machines in was so big, you could not see the end of the isle.
I have my doubts about plant grounding. Only a deep ground that hits the water table will be a real earth ground. Anything less that this is hit or miss.
Inverters and motor contactors generate huge amounts of noise. Less than perfect grounds will cause no end to issues.
In these cases, do you stop the installation and force the plant to fix their ground? I have managed this a few times, but it is an uphill battle.
In many cases, connecting any instrumentation to the plant ground can cause more problems than it fixes.
After a few big arguments with plant management, I have simply installed my own instrument ground and used for the installation. When this fixed the issues, we charged them for it and moved on. Let them argue about plant grounding with themselves.
Working system trumps all the "rule of thumbs". Recently, I have been supporting automotive assembly systems using PLCs and robots. They don't seem to have many issues with noise, except for the IT side and the Ethernet communications. But that is usually just a bad wire or several bad connectors in the system, or when someone accidentally connects the Ethernet back into itself as a loop. But that's to be expected. I think the industrial robots and AB and Siemens PLCs are designed as robust industrial systems that are less sensitive to noise issues compared to the instrumentation/ controls I used to support.