I notice that when I (on Friendica) respond to a Mastodon post with a content warning, I get an <abstract> element, and it certainly doesn't act as a content warning. So there seems to be some inconsistency in implementation.
That said, I'm not getting a content warning from this post.
A sensitive flag sounds like a great idea, but repurposing a summary as content warning based on a flag sounds like a bad idea to me. If there's no explicit content warning field, I think a spoiler tag is a better fit, because that already functions similarly without the flag.
But I'd prefer an explicit content warning field. There's clearly a need for it, and if it's explicit, you could set your client to always show or hide certain content.