I'd guess that this depends on the field. In mathematics in the U.S., one factor is your relationship with the student. If you are the Ph.D. advisor or a close mentor, then the letter should be at least two pages long (maybe rounding up a little) or it risks looking unsupportive, and three or four pages is not uncommon. For other letter writers, I'd recommend at least a page, and good letters are often two pages or longer. Of course these aren't absolute rules, and it's possible to write a great letter that's shorter, but it's a little risky.
The reason for the length is to give a chance to discuss the student's work (you can explain what they did, what some of the background and context is, why it's exciting, why the student's contributions are impressive, etc.). You can easily fill most of a page with generalities about other things - what your relationship with the student is, what your general opinion is, a paragraph about teaching and communication skills, etc. It's hard to cover this and say much about the student's work within one page, so a one-page letter sometimes suggests that either the recommender doesn't know the student very well or they don't have much to say about the work.
Overall, I'd say that whatever is needed is OK. The worst case scenario is that the committee will start skimming partway through, but that's not a disaster.