No, the huge numbers are perfectly valid to define an almost-straight arc.
The problem is that LinuxCNC is seeing a start and end point of the arc with such a small angle between them that it assumes that the start and end are identical, and that a full-circle is required.
The underlying cause is assuming that a very small angle is enough of a check, whereas there should also be a check that the distance between the start and end is very, very tiny too. (it's really a check that they are identical, with some allowance for digits of precision)
Fixing the actual tool motion was fairly easy, but it took me longer to fix the graphical preview, which uses different logic.
