I suspect that using underlining wouldn’t be enough to stop scams like this one; even if most people spot it, it only takes one inattentive person for a slightly-smaller-than-quorum cabal to win the game.

An advantage of underlining over italics is that it’s unlikely to happen by mistake, at least. (The rule does, however, need to be worded in such a way that links – which are underlined in some browsers – don’t count as flavour text, because we have some links in the rules that are meant to be actual rule text.)

eh seems too risky to me. feels like its just begging to be scammed somehow

If you define examples as flavour text, doesn’t that remove all legal force from them?

If the literal wording of a rule plainly said X but the example said the opposite, I assume we’d just go with what the rule said rather than regarding the situation as having any genuine ambiguity. The same way that if a rule contradicted how the proposer informally described it during the proposal discussion (without making any of those clarifications rule text), we’d only go by what the rule actually said rather than what the player said they meant.

Examples are fine as ruletext, if written clearly.

(I mean if the underlined example said opposite, there. If a rule said X in plain text and then gave a plaintext example that contradicted it, there’d be some scope for arguing what the effective rule actually was: if the contradiction was flavour text which “is not considered to have any meaning beyond being a string of characters”, that scope disappears.)