Having struggled with the Grounding debate myself (Land Rig), I have come to the following conclusion:
Grounding and bonding should not be looked at as separate options but as complementary, two sides of the same coin. Bonding providing the fault return path directly to the source for grounded systems and a fault return path between two grounded phases of an ungrounded system, and Grounding maintaining the potential between earth and the equipment during all conditions, as well as providing a known path to earth, reducing potential to ground during lightning strikes and inadvertent high line contact (obviously this also requires good bonding).
When using grounded and ungrounded electrical systems on a single unit I find it difficult to justify bonding only, for the above reasons. It appears that most people discussing this topic agree that good bonding is essential so in the following paragraphs I will look at grounding requirements only.
The NEC (in my opinion) is clear on the requirement for grounding on grounded (250.4(A)(1) + (2)) and ungrounded (250.4(B)(1)) systems, each of the three section specifying a connection to earth, the NEC later goes on to clarify this requirements and how they may be achieved.
A common example pertaining to my own situation (Land Rig) is 250.52(A)(5) where it defines the use of rod electrodes not less than 2.5m in length and in 250.56 which requires an additional rod electrode if the first does not produce a resistance to ground of less than 25ohms (which it does not on many land rigs).
Clearly for safety reasons the NEC is specifying two rods, 2.5m deep as a minimum, I would suggest additional rods possibly including a ground ring for redundancy or over capacity due to the harsh working environment and the portability of the equipment.
The NEC itself is a safety based set of rules designed to allow minimum safety standards to be reached, it is not a set of job specific technical requirements, with design requirements often being substantially greater than the NEC minimum for technical reasons as well as providing for redundancy / overcapacity. Indeed to lower the resistance to ground further Deep rod grounding is now in vogue with numerous reports extolling its virtues, these are available via simple internet search.
To argue the necessity for grounding is healthy debate, to ignore the NEC (read minimum) guidelines is misguided.
The above of course is open to debate and any suggestions/thoughts would be appreciated.