This is crossposted from the Thumper thread in the regular forum - it was in response to David's post about the algorithm that Thumper shot's are using to detect hits.  I'm crossposting it as it probably is more appropriate here, since it's a bug rather than a balance comment.

The short version here is that there is a bug with the algorithm that will result in a substantial number of Thumper hits going undetected when the Creep is moving closer to the center of the Thumper.

Details, reproduction instructions and the full crosspost below:

I'm afraid there actually is a bug in the code that you posted - or not so much in the code, but in the algorithm.  What's happening here is a collision test between a moving point and an expanding donut (creep centroid and donut area), and the problem is that it's not robustly dealing with a special case - that if you're simply checking overlap, the objects may pass through each other entirely between overlap checks and never trigger a "hit".

Imagine the following situation, using generic units:

Frame N: The current radius of the ThumpShot is 1, and the previous radius is .5   The Creep's centroid is currently at radius 1.1, but is heading towards the thumper at .15 radii/frame.  Since the Creep is not within the .5-1 radius interval, there is no hit.

Frame N+1:  The current radius of the ThumpShot is now 1.5, and the previous radius is 1.  The Creep's centroid is now at radius .95.  Since the Creep is not within the 1-1.5 interval, there is no hit.

Key:  Black dot is the centroid of a creep.  "Donut" with red stripes is the donut that is currently getting checked for overlap.  I would have made the stripes purple, but I don't like jelly donuts.  Note that the centroid overlaps the donut in neither instance.

This is likely to occur especially often with speedy creeps, but in theory could happen to nearly any Creep that ends up just outside the radius in one frame - even a small movement closer to the center can take it within the radius of the last ThumpShot.

To verify that this is happening, I did a test on "Down The Chute", configuring the starting grid with just a single thumper as follows:  From the last acceleration pad by the entrance chute, I placed the thumper two hexes out in the direction of the bottom-right edge.  This results in a situation where the creeps are traveling towards the center of the thumper (and hence decreasing their radius) for nearly an entire hex (they're out of range for part of the time).  I also placed a bunch of towers near the exit to buy a bit more time to watch.

If you watch this situation for several waves, you should see several Creeps that are clearly having their centroid pass through the wave which are escaping without a scratch.  Since the bug is timing dependent, it could potentially take a few tries to encounter this (though it actually seemed to be occuring on a relatively high % of hits in which the creep was approaching the Thumper's center).  It's actually easiest to spot on the slower moving creeps, as they spend more time approaching the thumper's center.

Since roughly 50% of Thumper hits should be on creeps moving toward it rather than away from it, depending on how often this bug is hit (and unfortunately, it's being hit pretty often from my non-scientific observations) it could be severely negatively impacting Thumper damage.  I would be willing to bet that it's also responsible for some of the observations of "Thumper Wave Nullification". 

EDIT:  I just noticed  that this also happens in reverse - occasionally creeps will get double hit while moving away from the Thumper.  It's even remotely possible that they could be triple hit.  In the test case I'm using, in addition to seeing some misses that shouldn't occur, you should also see some creeps being one-hit by the thumper when they otherwise should be able to survive the hit.  This means that on the whole, this bug may not actually be adversely impacting Thumper damage output - but at the same time, it's certainly responsible for some of the perceived unreliability of the tower.  Generally people are going to notice the tower whiffing and not notice the tower hitting twice.

EDIT:  Here's how I set the board up for my test.  Remember that since this is a timing based issue, it may take a while to clearly observe a miss.  I find that the first wave of slow movers is the easiest to spot a miss on (since it's usually more clear that a miss wasn't due to them moving out of range too quickly to be hit)