Indeed - the so-called "shy Tory syndrome".
Over on politicalbetting.com there is a lot of discussion of the methodology of the various pollsters and how they deal with (or don't deal with) those who say don't know. The pollsters that reallocate typically show smaller leads as the reallocation is largely to the tories from 2019 tory voters.
But I think pollsters have a real problem if they baseline everything on the 2019 election as that was such a weird, one issue (plus Boris vs Corbyn) election. So to presume that 2019 tory voters are somehow 'historically' tory and even though they may say 'don't know' at the moment many will fold back into the tories seems a tad naive. We know that a considerable number of 2019 tory voters were either first time tories or were typical non-voters. Hence the red wall falling. Will these voters 'fold back' into voting tory (as they did in 2019) or 'fold back' into not voting or voting labour as they likely did in 2017, 2015 etc?
To my mind 2017 is maybe a better baseline to use as a starting point for consideration of the 2024 predictions than 2019.