I think you have confusion over what's being grounded, or why.
(Image source: Texas Instruments - AN-1469 PHYTER Design & Layout Guide SNLA079D)
There are three commons (not necessarily "ground") in a typical Ethernet connection: PHY / digital / main circuit ground (show above as the three-line "ground" symbol); chassis or shield ground (shown as the rake-like "earth" symbol); and the Bob Smith termination "ground" (common) formed by the resistors.
The capacitor in question is only used to connect the BS common to an appropriate sink for ESD current. For equipment in a metal chassis, this will most likely be the chassis itself. For plastic enclosures, this will probably be the PCB ground plane.
Note the above shows 0-ohm jumpers, which can be removed if circuit-chassis isolation is required. In that case, most likely the 1500pF capacitor should be returned to chassis, not circuit GND. (With the jumpers in, obviously it doesn't much matter.)
The purpose is to close the loop as short as is possible. ESD comes in from outside, not necessarily on wires at all, but also as displacement current to nearby metal (chassis, ground plane, cabling), or waves entirely. The shortest loop you can close on that, is back to the same things: chassis, ground plane, cabling. If circuit ground is inside a chassis, returning the capacitor to that would needlessly invite ESD inside the enclosure, partially defeating the purpose of the enclosure(!).
Another point against returning the cap to DGND: since you mention "digital ground" -- I assume making the distinction to general circuit, signal or power ground -- that implies DGND may be noisy (due to digital signals running across/into it). While there is a separate conversation to be had about the merits of such a segmented-ground design, the relevant aspect here would concern noise on the ground with respect to chassis or cabling. Point in case, putting a capacitor here would literally couple noise from that plane directly into external connections, making a big emissions hazard.
As for ferrite beads -- you might use one when you need DC grounding between circuit and chassis grounds, but need AC isolation, or some impedance between them anyway (as opposed to a very low impedance from jumpers, or direct bonding). This might be relevant when you have other connections, daughterboards say, in a larger system, and there's some amount of common mode noise sloshing around between them, and no single board really has a good "ground" basis with respect to the enclosure around it. But this probably isn't a very common situation, because all those cards will be dropping DC across the same connections, so the common mode isn't limited to AC alone. And a few ~mV of DC will be enough to saturate a ferrite bead (again giving a low impedance); that is, if the FB itself doesn't melt from excess current! Generally, multiple current return paths is not a good plan, especially in systems like that. Keep loops small, both at AC and DC. In that case, probably an R+C network is all the bypassing you'd want to use, between circuit and chassis grounds, when you need to make a connection between them at all.