Speaking descriptively not prescriptively:
(1) Re the upper- and lowercase forms for archaic numerals, although of course all characters on inscriptions (and presumably seals) are technically uppercase, editors may choose to render then in upper- or lowercase depending on editorial preference, and in some cases this difference is even meaningful (lowercase representing units, uppercase thousands, for example). Unicode merely reflects this practice. While I woulod strongly recommend anybody using these to be consistent within their project/corpus, I don't think it is possible to impose consistency across the field.
(2) Re the two koppas (Q-form and lightningbolt-form, for want of better terms), my understanding is that the Q-form is the archaic letter and the numeral 90 throughout antiquity. At some point (I think fairly recently) the lightningbolt-form replaced the numeral and is now so used in modern Greek chapter numbers and the like. I don't know if the two are ever used in the same text with different meanings, but Unicode has clearly decided to encode the two as separate codepoints. I would suggest that when encoding ancient Greek texts, we always use the Q-form, whatever the actual shape of the glyph on the stone/seal/page.
(3) I presume there's a similar situation involving the stigma and F-shaped digamma, when used as numerals; the latter is also sometimes an alphabetic character. The difference being that I believe most people having encoded numerical digamma=6 using the stigma codepoint.
(see also Greek Fonts (variant character forms))