Sorry for the hesitancy in an answer to the question. The subject is long, difficult and complex. In Other Words, it requires painful thought – I HATE when that happens.
There is the temptation to use the old standby “Because it’s in the code”. In the case of the 1999 NEC (Because that’s what Virginia uses), it is in 250-92 (a)(3).
In order to understand the principle we must remember, and understand, the basic concept of the ground. Basically stated:
The concept of the ground is to force (give the current clear path) fault current to a high enough level to clear the fault, or open the OC device in a timely fashion. This is most often defined as 5 cycles.
The explanation of that principle normally takes me 6 – 12 hours in class. Let’s face it, Bill would bar me from the board for that. If you do not accept that as the basic principle of grounding, then no explanation can satisfy the question.
Do you accept the concept of transformers? That power applied t o the primary winding “magically” makes power appear on the secondary winding. There is no connection between the windings, but merely by inductive effect the secondary is drawing (using, wasting) power from the primary, regardless if anything is connected to it.
Transformers have a high efficiency because of laminated iron/steel cores, winding size and placement, insulating material, etc.
How efficient is this principle when it is a conductor (bare or insulated) laying in a metallic cover? In this case – we have an energized conductor (during a fault) and by the same inductive effect it has energized it’s metallic covering during a period we want the current at it’s highest level to open the OC device. Is it attached to a conductive surface, such as a steel bar joist or column/beam, or even concrete? Then we’ve energized that portion as well again robbing us from our fault clearing power.
The conduit bonded together at each end to the Grounding Electrode Conductor is a clear parallel path, not bonded at either end it has the same sympathetic relationship that transformer windings have, bond it at only one end and we have a series circuit with sympathetic values at the unbonded end. The series circuit configuration will not only draw power from your critical fault path but it will arc internally and likely throw some very hot sparks on the path – every time you have a fault, large or small. A simple fault on a branch circuit upstream places a current value back here as well.
What are the sparks? You’re losing conductor and/or conduit every time you fault or current goes high, how does this effect your future performance? Let’s say it does not improve your values, or performance.
The 7th edition of “Soares Book on Grounding”, on page 107, Chapter 7, finishes their explanation this way “Where that bonding procedure is not followed, the impedance of the grounding electrode conductor is approximately doubled with the result that it’s effectiveness is markedly reduced”.
That’s about it in a nutshell, hope it helps. As always, I strongly suggest “Soares Book on Grounding” for a deeper explanation, with pictures.
BTW, 1959 NEC, 250-71 requires the same thing, says it a little bit differently, but actually makes it more clear. This is NOT a new requirement.
Like I said earlier, “Because the Code says so” and has for a LONG time.