Richard, while I generally agree with your "rants" on this topic, I will say that engineering calculations are often (relative term, I suppose) done by hand where you round each variable as you go. The final answer is typically rounded up to the next standard size, so the loss of precision does not affect the final answer.
Whether the loss of precision affects the final answer depends on how much precision you use and the need for accuracy in the answer. Four significant digits may be enough to calculate the required trajectory to put a basketball through the hoop from the center of the court, but it's not going to be enough to calculate the trajectory of a spacecraft destined for Mars. If the calculations are done by hand the engineer (hopefully!) knows that though, and uses enough precision to solve the problem at hand to the required accuracy. The danger of the float keyword is that the engineer may believe they are using enough precision because they don't realize what it is doing, and that the answer therefore does have the required accuracy, but they are wrong. And so the spaceship crashes into Mars (something that is even more likely if calculations are done outside of Mathcad, with no automatic unit handling Mars Climate Orbiter Team Finds Likely Cause Of Loss 
). That's what drives me to try and shut down this suggestion every time I see it come up.
It's cumbersome, but this is likely the only workaround that does not result in loss of precision. As noted above, you can replace "float" with "round" or "Round":
Obviously, you'd want to put most of these calculations in hidden regions.
