A few things
1) The question of drag is, first and foremost, a gameplay question. As others have pointed out, a game in which speeding up and then coming to a stop takes "realistic" times is....probably not one that I want to play. On the other hand, if you can speed up and slow down in a "reasonable" time, without a drag force, it also implies that you can reach insanely fast speeds in a reasonable time. Frankly, I think it makes for bad gameplay. That's a personal opinion, nothing more and nothing less
2) Limited top speed without drag is, I would argue,
even less Newtonian than having a drag force. There's nothing non-Newtonian about a drag force. True, there's no medium to exert drag in space. But I think it would be better to call it "unrealistic" instead of "non-Newtonian," which somehow suggests that the real "Newtonian" physics uses some kind of advanced internal machinery, when in reality,
both methods will always use discrete Newtonian integration. Just that one applies a drag force and the other doesn't. Interestingly, a hard cap on the velocity without drag
does not fit into the framework of Newtonian mechanics (please correct me if I am mistaken), because it cannot be implemented with a force! Velocity cap without drag suggests either a hack or relativistic mechanics, both of which are highly-non-Newtonian
3) Although it's not the major issue, there
are some technical problems with unlimited-top-speed systems. Precision, as others have mentioned, will be problematic. But mitigated substantially by double precision. A bigger issue, though, is collision detection. You simply won't be able to perform accurate collision detection in a timely manner with unbounded velocities. It will become either too inaccurate or too expensive. If you're willing to sacrifice CD accuracy, then my guess is that you wouldn't have too much trouble implementing a game with unlimited top speeds. Of course, at some point, you need to cap the velocity. But I suppose if you use real relativistic equations then you get a speed cap automatically.
4) BONUS ROUND! You mentioned a zero at the bottom of an equation. Which is an interesting thought. Incidentally, if you set the drag coefficient in the LT engine to be zero, you
will indeed get several zero denominators in various places and cause things to blow up. The reason is that any piece of code or equation that expects a finite top speed will blow up in the absence of any drag. So the AI, for example, is using it's knowledge of the drag force to make more accurate predictions of where objects will be when they come to a stop. If drag is set to zero, an object in motion will never come to a stop unless acted upon externally, so the AI would totally fail. It would actually require a different steering algorithm than the one LT uses to manage AI pilots in a zero-drag, no-speed-limit setting!