Sorry for the late review, but these three cases don't actually trigger FURB129, so I'm not sure if they're necessary. I noticed this while rebasing the 0.12 branch.

I think this is sort of marginal -- I see some value / little cost in having these kinds of pathological cases here just to make sure they don't cause a panic, but taking that attitude to the extreme is obviously untenable :) I can remove since I'll be doing a follow-up PR anyway, but if I weren't changing the other piece in readlines_in_for.rs, I'd probably opt to just leave these here.

Makes sense, I can see the value too! I was just afraid I dropped a few diagnostics while rebasing until I realized they weren't meant to trigger. I'm happy either way in both cases here.

While I'm here, I also think this section would be a bit nicer as:

    let deletion_range = if let Some(parenthesized_range) = parenthesized_range(
        expr_attr.value.as_ref().into(),
        expr_attr.into(),
        checker.comment_ranges(),
        checker.source(),
    ) {
        expr_call.range().add_start(parenthesized_range.len())
    } else {
        expr_call.range().add_start(expr_attr.value.range().len())
    };

    let padded = pad_end(String::new(), deletion_range.end(), checker.locator());
    let edit = if padded.is_empty() {
        Edit::range_deletion(deletion_range)
    } else {
        Edit::range_replacement(padded, deletion_range)
    };

Will change!