Without a ground connection there is potential (excuse the pun) of leakage or wiring fault (or rarely a component fault) biasing the circuit voltage; this is bad because DC control wiring is nominally SELV (safe extra low voltage) and many safety provisions and even component ratings depend on this. I once encountered a problem with unbonded ring wiring (UK) where one phase of 230 AC power was biased to 640 V above ground (exceeded the insulation rating of some equipment). BTW, SELV also has a power dissipation limit which is either 240 or 480 VA in various standards. This grounding should be equipotential (a.k.a. single point ground); if it produces a ground loop, you're doing something wrong.
However, I have noted that several popular field buses communicate ground even beyond a single control zone which is a SAFETY VIOLATION but may also be victimized by a ground bond test. Ground bonding of DC control circuits is supposed to provide a level of safety and is required or recommended by many safety standards. One caveat is that some DC power supply manufacturers only rate the safety of their device with the exception of the AC power connection (check their compliance declaration and don't buy).
Shield and ground though at a common potential are two different things and should not be intermixed. Use of shields to communicate ground potential is generally wrong. I once spent nearly a year looking for a ground induction fault in a high quality audio circuit until it was discovered that a shield foil in an isolation transformer was connected to chassis ground (not analog ground). Suffice it to say a proper shield should not be capable of communicating a signal. A significant issue in control systems is the connection of servo amplifiers where the typical manufacturer's instruction is to bond the motor cable shield to the panel which in addition to dumping a lot of noise into the ground plane (particularly with PWM drives) as well as creating the risk of an induction loop if the motor case is grounded by physical attachment to the machine frame; there is no great solution for this although making one corner of the cabinet back-plate the equipotential ground bond point helps. The other place where grounding frequently goes south is in the wiring of power line isolators: even though manufacturer's instructions clearly define an input ground and an output ground and a minimum separation distance (usually at least as large as the length of the isolator) it is not uncommon to find that it isn't wired that way or that only one ground wire is attached (?!); this gets even more complicated when the standard requires that the supply ground lead be directly attached to the chassis, however, improper grounding of a power line isolator turns it into an expensive hood ornament. When connecting complex controls with sensitive signals ('sensitive' being a relative i.e. level of noise versus sensitivity of devices) a ground is ground philosophy is wishing for trouble. In my experience, the vast majority of stability and lost position problems with servo systems have been the result of ground circuit problems. Ground IS A Circuit.