I don’t know, it could probably work, America is the outlier for their election seasons. UK elections are held 5-6 weeks after Parliament is dissolved. The 2022 French Presidential election was held less than 2 weeks after the polling date was announced. Comparatively, the USA’s 7 months to convention, 10 months to election is a lifetime. You can do loads in 106 days
The problem with exit polling, as with the problem with polling in general (exacerbated by the modern age), is that they’re voluntary. The simplest explanation is that a higher percentage of women answered the exit poll than men. Or that women who voted for Trump were less likely to answer the poll. Or the people lied when they answered the poll.
There can also be statistics reasons for it too. Not knowing the methodology behind how this was collected, but you can also have selection effects. If I’m trying to run a statistical analysis on a population, I want as many respondents as possible to reduce the error and deviation, but I also have to operate with limited funds. Be much more efficient to post a few people up in higher density places like cities that tend to vote more blue anyways than having pollsters scour the backroads of Wyoming, for example, where I would wager a higher percentage of women voted for Trump.
In the end, don’t put too much stock in pre-election polls, and definitely don’t put too much stock in exit polls. Think about it like this, if you got a phone call from a random number, would you pick up and answer questions about how you vote in such a controversial election? If the answer is no, then you know why polls aren’t accurate