Score currently seems to be mostly representative of how many questions they've done, not how "good" someone is at getting questions correct (eg. it's more like a sum, rather than an elo-type ranking).
I think if we make penalties 50 instead of 30, it'll help alleviate some of this? Alternatively, we could weight questions differently based on what percent of people have gotten that question right so far, although this is definitely more complex and unintuitive.