Diff rendering assumptions #20080
Description
I only just realized that our fix infrastructure always assumes that all changes belong to the file from the primary span.
How would the rendered diagnostic look like if we had a sub-diagnostic with a span pointing to a different file? I don't think it's something we have to solve now but it might be worth documenting if the rendered output looks confusing.
Originally posted by @MichaReiser in #20036 (comment)
Beyond the implicit fix-primary-span connection, we'll also render diffs immediately after the last diagnostic, whether that fits nicely or not. Especially after the planned changes in #19919, the diffs will be expecting a help: header above them, which will be missing if an additional sub-diagnostic has been added before the help message.
These issues shouldn't affect any existing diagnostics yet, but they're something to look out for as we start using the new diagnostic features more extensively.