Unrotate pages rotated by failed text recognition

I do OCR on scanned images:

• Document Language: English (UK)

• Output: Searchable Image

• Downsample To: 300 dpi

This generally works fine. However, there are two pages in a document of 220 pages where Acrobat:

1. fails to recognize text,
2. it rotates the scan quite a lot.

For point number 1, it is forgiven: There is not much text, and that text is stylized. I don't even expect it to recognize that.

However, point number 2 is a big problem. How do I undo these rotations?

I tried to manually rotate with the edit tool. However, this can only be done in 2° or 1° increments, and it's not possible to rotate the image back to its original position.