Removed pages are gone, not hidden
This matters more than it sounds. Some ways of "removing" a page only stop it being displayed — the page object stays in the file, and anyone who opens it in the right tool can recover the content. That is how confidential material ends up in documents that were supposedly redacted.
What happens here is different: the pages you keep are copied into a brand new document and the ones you deleted are never copied at all. There is no leftover object to recover, because the output file was built from the survivors rather than edited down from the original.
That does mean the operation is a rebuild rather than a patch, so the saved file is a new file. Bookmarks pointing at deleted pages disappear with them, which is the behaviour you want.
Keep only, when you want three pages out of ninety
Selecting eighty-seven pages to delete is a miserable way to spend a morning. Select the three you want instead and press Keep only — the result is identical and the selection takes seconds. It is the same operation viewed from the other end, and it is the one most people actually need when pulling an appendix out of a report.
Selection works the way it does everywhere else: click a page, shift-click to extend a run, hold Ctrl or Cmd to add individual pages. Ctrl+A takes the whole document, and the toolbar always shows how many pages are currently selected so you can check before committing.
Nothing is final until you download
Every action here — delete, keep, rotate, duplicate, reorder — is recorded in an undo stack that goes back thirty steps. Ctrl+Z reverses the last one and Ctrl+Shift+Z puts it back. Reset returns the document to the state it was in when you opened it.
None of it touches the file on your disk. The original sits where it always was, untouched, until you download a new copy — so a mis-click costs you a keystroke rather than a document. This is also why there is no "are you sure" dialogue on delete: the undo is cheaper than the interruption.
Before you send the result to someone
Two things worth checking. First, if the pages you removed were sensitive, remember that any copy you have already shared still contains them — removing a page from your copy cannot recall a copy sent yesterday. Second, if the point is to hide part of a page rather than a whole page, deleting pages is the wrong tool: covering text with a black box leaves the text underneath perfectly selectable, and proper redaction has to remove the underlying content.
For the common cases — dropping the fax cover sheet, removing the blank separator pages a feeder inserted, sending someone only the section that concerns them — removing pages is exactly right, and the output is a clean document with no trace of what was cut.