Offshore oddsmakers made a killing in 2016 too! I think it's a bit too easy to describe people as irrational and gullible for being willing to risk some cash on a prediction. If it was irrational to hope that reelection was possible, then it was irrational to worry that reelection was possible. I managed to lose money on PredictIt in both 2016 and 2020 by betting on what I thought the more likely outcome would be, both times struggling to weigh the forecasts of others. It's still an open question as to whether prediction markets are more reliable than the alternatives.
A rational market requires perfect information and perfect rules. The more imperfect the information and rules, the less rational the market will be. This is the whole point of high frequency trading: if you can access the information faster than the next guy, the imperfection of information flow favors your participation in the market. TSLA is heavily overvalued at the moment. The classical economic viewpoint on this is that TSLA will sink in value because its performance does not match its metrics, or the market metrics do not properly capture its performance. There is no aspect of economic theory to describe the millions of hipster nerds that want TSLA to be worth more than Ford and Chevy combined. Economics presumes charity and philanthropy to be rationally motivated because economic theory is largely bullshit. There is no room in the model for the idea that billions of dollars fucking hates Chevrolet and everything about it and is willing to burn in order to disrupt the status quo. That zeal distorts the economics of TSLA and allows them access to easy financing even while losing money for years at a time. AMZN would have gone bankrupt by now if it weren't for the earnest love of managers and investors not only discounting the possibility of anti-trust but actively longing for monopoly. You lost money on Predictit by thinking the inputs and outputs have a correlation. This is an article about why they don't. "all models are inaccurate but some are useful" - in this case, this model is useful for the House.
No doubt there is some wishful thinking, and claims of betting may just be attempts at market manipulation, or trolling. The probability is non-zero until it's zero. PredictIt called my position on December 18, four days after the electoral college vote. In 2016, faithless electors hurt Clinton more than Trump, so that vote was not a formality. January 6 is the next opportunity for drama. If the betting markets work efficiently, it should not be irrational to bet for either side to win; the payoff for making a long-shot bet should be high enough to balance the risk of losing.