Since this bug has been with us in the code for a long, long time, I'm reluctant to just fix it in a point release, because this will doubtless change the effect of a number of user scripts.

However, while the code doesn't agree with the documentation, updating the documentation seems wrong too:

1. having -daystart documenting as setting the origin at the beginning of tomorrow seems deeply unintuitive.

2. it looks to me like the current code doesn't align the origin with midnight around the points when the clocks change, either, though I have not verified this in testing.

There's no obvious answer which fixes the bug but won't break scripts.   However, here's one idea for a transition.   Comments appreciated.

1. Modify the existing -daystart so that it takes an optional argument.   The argument can be "tomorrow-ish" or "today".   "-daystart tomorrow-ish" has the same meaning as -daystart does now.    Document the precise effect of tomorrow-ish.   ("ish" because of the problem with short/long days).  The argument "today" makes find do what the documentation says it should.   If no argument is used, a warning message is issued and "tomorrow-ish" is assumed.

2. Either deprecate "-daystart" without a following option, or warn that its meaning will change.

Comments appreciated.