Every car you’ll drive with a fuel economy gauge behaves this way. When you coast above a certain speed/rpm Your mpg will be more than 99.9. Upon slowing below the point where idle fuel consumption (coasting in gear, engine turning, though not being fed any fuel) results in less than 99 mpg you will again see the mpg drop, indicating fuel actually IS being fed into the motor, albeit very little, reaching 0 when you’ve stopped. Every car does this.
Just as the CVT gear ratio gradually changes during acceleration, it changes during deceleration, though not exactly in the same manner.
I'm not sure I entirely follow what you're saying here: "Upon slowing below the point where idle fuel consumption (coasting in gear, engine turning, though not being fed any fuel) results in less than 99 mpg..."
If the engine is not being fed any fuel, then fuel consumed is zero and the distance traveled per gallon consumed is infinite* as long as you're moving at all, no matter how slowly. If the engine is fed enough fuel to idle, then it
is being fed a non-zero amount of fuel, and there is a speed where miles per gallon falls below 99.9.
In my case, when I was looking into this, the reported instantaneous MPG would stay at 99.9 until it dropped abruptly (not gradually) to the 60s or so (along with a very subtle shudder), and then declined gradually from there as the car slowed. This suggests that the car is being fed no fuel until a certain point, after which the fuel flow is switched on and fuel consumption became a finite number (and well under 100 MPG).
* Some mathematicians may quibble about this statement because it involves the result of division by zero, and a non-zero number divided by zero is undefined in formal math usage - they are happier saying something like "for non-zero X, the limit of X / Y approaches infinity as Y approaches zero". As a practical matter, however, scientists and engineers are perfectly happy to avoid having to handle special cases separately, so they are willing to say "... so we will treat X / Y for non-zero X and Y = 0 as infinity", because doing so is actually meaningful to them.
For a concrete example, look at how the IEEE Standard 794 for Floating-Point arithmetic handles this - the default option for non-zero X divided by zero is to return the floating-point representation for infinity, and keep going. There is an option to return NaN (Not a Number), and either keep going and let the meaningless result propagate, producing NaN for any result subsequently involving the NaN ("Quiet NaN"), or force the program to either stop with an error or "trap" and handle the special case ("Signaling NaN"). As far as I know, the default is typically used, along with the converse, "any finite number divided by infinity is zero".
The result of zero divided by zero is universally agreed to be undefined. In IEEE 794 math it results in NaN, I think "Signaling" by default, which forces the program to explicitly deal with it (or abort). Infinity divided by infinity is also undefined.