Blobistat. Though it's not a tower defense game, it is an evolution simulation in which the player is tasked with the goal of poisoning all the predators (germs) before they eat their pink, helpless prey. Each germ has a unique resistance "genotype" which is displayed in a white bubble, signifying the minimum concentration of each antibiotic needed to affect it. The 4 antibiotics do different things: slowomycin reduces speed, nomifloxin prevents them from eating, stericilin prevents reproduction, and necrodizole, the most expensive, depletes health. As the germs reproduce, their offspring get "mutated" versions of their genome, so that the trend in resistance statistics changes over time depending on which germs survive.
While playing it, I did feel a sense of panic when a germ ambled toward one of the prey objects. But the "eating" itself passes almost unnoticed (a little blood is in order). This is just one example of how this project lacks satisfying visual, auditory, and tactile feedback.
Maybe a bigger issue is that it's poorly tuned. The game has only two stable states, and so it's easy to run out of antibiotics and watch the game field become totally overrun with germs. It is actually the complete opposite of adaptive difficulty.
There is also the problem of what happens when you run out of points to buy antibiotics - since points are awarded every second, but the antibiotics are purchased each frame, there is a strong "popping" effect that does very little damage to the germs and also fails to communicate bankruptcy to the player.
I think what Vincent really means is not that the model itself is boring, but that the game is boring. We need to work on that.