(given a pool of 100, not even 32!)
Try looking at the cumulative probability for P(X>=2) when you manipulate the numbers a bit. Even if you just change the base probability of mortality to 1%, it jumps to 26% chance. If you restrict the population to 32 people, the threshold for it be more likely to happen than not is a 5.2% chance of death.
Plugging that in the binomial calculator, P(X>=2) at 32 trials is... 0.09%!! Astronomically low odds.
And you are right, their deaths are certainly not independent :)
---
Probability = 0.01
Number of Trials = 32
Number of Successes = 2
Now you look at cumulative probability (P X>=2) = 4%
---
You are correct that if bump the chances up to an annual 5.2% of death, then finally it starts becoming reasonably probable. So, for contrast, whistle blowers at Boeing seem to have approximately the same survival rate as somebody diagnosed with stage 3 breast cancer.