Hallo Sanjeev,

first, I fully agree with Erik's statements! Especially with:

The amount is typically related to the age of the plant and, of course, to what is going on in the plant.

How can an estimate formed about amount of common mode noise (CMN)? This is indeed a difficult task. It's the best to distinguish the relevant sources of CMN. Of course, only a few can be discussed here, and only briefly:

1. Ground loops: Is ground routing of plant or system not done, that ground mesh or grid is formed, connections between certain ground points introduce ground loops. It makes no difference how these connections are made, either by metallic things of plant, like frames, pipes, stairs, or by yellow-green earth cable from mains, which is connected to 0V of circuit in devices which must obey safety class 1 rules.
If now a varying magnetic field intersects these ground loops, a voltage is induced like in a transformer winding. There are two ways to decrease this induced voltage, either magnetically shielding sources of heavy magnetic fields, like motors, generators, transformers, etc., by just putting it into a non-alloyed steel enclosure, or by introducing a big number of additional ground loops, finally forming a ground mesh or grid. By this, loop areas become smaller and smaller, and induced voltages will cancel each other, finally resulting in a system with drastically decreased CMN. Ground mesh represents an idealized ground plane, where no voltage can be induced, as we know.
Amount of induced voltage is often overestimated. If you bend a loop arround powered 5VA mains transformer with field direction perpendicular to loop area and measure induced voltage by a digital multimeter, you will not find anything bigger than 1mVeff. So, only if magnetic fields come from much bigger sources, CMN will become relevant.

2. Lightnings: If a factory is hit by a lightning, directly or indirectly, a heavy current will flow along ground routing. If it's not well earthed, like in an old plant, impedance of ground routing is high and relevant voltage drop occur. So, between two ground points CMN is introduced. Here also, ground mesh technique helps in drastically reducing CMN.

3. Surges on mains voltage, which are shunted via protection circuits to ground: The same like (2)

4. Often overlooked, interwinding- or stray-capacitance of mains transformer: Assume having two devices separately powered by mains transformer circuits. The one circuit's 0V is connected to yellow-green earth cable, the other does not see such a connection, but is floating. Unfortunately this is no ideal floating, because mains voltage is connected to 0V, via interwinding capacitance of mains transfomer! Interwinding capacitance is about 100pF for smallest transformers, but rises up to a few nano Farads for big toroidal mains transformers.
If now both circuits are connected to each other and symmetrical signal routing is used, a big CMN can be fabricated, if there's no direct connection between the two 0Vs, because it must not be connected for other reasons. This can be observed just in your apllication! Assume that your sensor circuitry is the one with floating 0V and '4-20mA to 0-5V' converter is the one with earthed 0V. Then a voltage divider is built of interwinding capacitance and common mode input impedance. In our example this could be 100pF in combination with 400kOhm / 2. At 50Hz impedance of 100pF is about 32MOhm, so that a CMN of about 1.44Veff (50Hz) appears at input of converter.
This seems to be quite small, but don't forget, that for interference of higher frequency impedance of interwinding capacitance drastically decreases and by this CMN at converter increases!
If interwinding capacitance is much higher, let's say 2nF, CMN (50Hz) increases to about 30Veff! Peak voltage then is greater than 40V, and although voltage appearing at non inverting input of converter is divided by 4, converter gets a problem, because input voltge begins to overrange.
Keep in mind, that not only interwinding capacitance can introduce capacitive coupling to mains voltage, but also common mode filter, when using a switcher design!


So, what does all this mean to you?
It's of high benefit, if you power converter and sensor front end by the same supply, means having an identical 0V reference.
If you can't do that, at least connect 10-100nF from floating 0V to earth, if possible. This drastically decreases CMN appearing at converter input.

What about shield?
As discussed above, shield can introduce a ground loop. This does not necessarily mean a disadvantage, if it's a part of ground mesh. But if no ground mesh technique is used, forming such a SINGLE ground loop can result in an desaster, like Terry reported about, some weeks ago.
There's no midway between ground mesh technique and floating applications. Either you connect 'everything to everything' and by this forming a ground mesh, or ground loops should be avoided, totally.
In former times floating technique was widely used, you know that stuff using opto-couplers and isolations amplifiers. But today, with aggressiv cellphone radiation and super high frequency transmissions, the much more high frequency suited ground mesh technique becomes more and more demanding. The consequence is, that today we have to deal with mixed systems many times, means high frequency applications in 'old plants'.

So, my advice to you: If two separate grounds are involved, do connect shield galvanically to ground at only one end, unless cell phone radiation and other high frequency interference makes big trouble. Connect it to the side, where filtering and protecting measures are more difficult to realize. So, many times I connect it to outputs.
If problems occur, also connect other end of shield via 10nF to local ground.

Good luck,
Kai