So this is "right-dislocation", where it looks (on the surface, at least) like some element of the sentence has been moved off to the end of the sentence, and it's certainly very colloquial and also certainly nowhere near marginal enough to be just ungrammatical performance error. The overall syntactic view at the moment seems to be that there are really two different constructions masquerading as one, where the difference is, basically, whether there's case marking—が, を, に, etc. (but not は or も!)—on the dislocated element. And the ones that you're talking about in your examples are the ones without case marking, so I'll talk about those first.
... I said just a moment ago that they weren't just ungrammatical error, but from the viewpoint where we have two different constructions going on, these ones with the bare dislocate are kind of... swept under the rug, syntactically. We mostly treat them as being fragments related to the main clause just by context and apposition. This doesn't have to mean that they're erroneous or unplanned; they could be following a productive lexical schema (e.g. 何 + Dem-NP), or they could be chosen for various pragmatic reasons. They're not really used to disambiguate between genuinely plausible alternatives—you only typically see them (outside of recognised formulas) when the fragment is already highly given—これ, それ, 私, etc.—but you're right that they're an anti-focused way of securing a reference that might not land 100% as clearly as it should, or sometimes of re-pointing at the referent after front-loading the emphatic part of the sentence. And another reason you might use them is just to lengthen the utterance, maybe to soften it a touch, or maybe to give the other person a bit more time to be ready for their turn. (Japanese, of course, devotes a good bit more explicit attention to turn-taking than other languages.)
The other kind, with overt case marking, is rarer and more involved. There seems to be more deep structure present—dislocations out of relative clauses for example are perfectly acceptable without the case marker, but become awkward to receive if the case marker is there:
"Tarou wants to know if Hanako has read it (the book Lectures on Government and Binding)."
太郎が花子が読んだかどうか知りがっているよ、LGB (fine)
太郎が花子が読んだかどうか知りがっているよ、?LGBを (weird!)
Also this form of right-dislocation permits negative polarity items such as しか, suggesting that the dislocated item is still in some deep structure with a negation in the local clause:
"Tarou hasn't read any of it (except Lectures on Government and Binding)."
There are various arguments about what's underlying all of this, ranging from "an element has moved rightwards syntactically" (extremely rare and maybe impossible in Japanese; people really don't like this) to "the sentence is syntactically normal but an element is being delayed at pronunciation time" (well, it explains a bit too much) to "the element is the only pronounced part of a full duplicate of the preceding sentence" (actually not that uncommon of an approach).
Regardless, the function of these is more like the genuine case where you're repairing the sentence with some piece of information you've just now decided was relevant or needed to be made explicit.
Which is a little bit strange just intuitively, right? You've got this weird syntactic fussiness in a post-hoc repair operation that you don't have when you're making a deliberate callback to something established in the conversation. But self-corrections are very frequently analysed as having this structure, of a fully built-up syntax that duplicates what you just said but is left mostly unpronounced. And really it's the case that, from the perspective of the brain, extra syntax is cheap, and context clues are expensive. So of course it'd reach for the former when you realise split-second that something has to be fixed, and for the latter when you're trying to go for deliberate social function.
