Requesting indexing moves a page up the crawl queue, but it does not decide whether the page gets indexed. That is the split the button’s reputation glosses over: it can get Google to come and look sooner, and that is genuinely useful, but it has no say over the indexing verdict once Google has looked. Treating the request as a fix confuses “crawl this sooner” with “index this,” and the second does not follow from the first.
For a page that is new, or one you have just meaningfully improved, the request earns its keep. It prompts a quicker crawl than waiting for Google to find the change on its own, so the page is evaluated sooner. What happens at that evaluation is unchanged by the request. If the page clears Google’s value bar, it gets indexed; if it does not, requesting indexing repeatedly will not push it in, because you have not changed what got it excluded. This is why hitting the button on a page stuck in “Crawled, currently not indexed” usually does nothing: the page was already crawled and judged not worth indexing, and asking again does not alter the judgment.
There is one wrinkle that explains the occasional false hope. A requested crawl sometimes gives a page a brief freshness lift that nudges it over the line into the index for a while, only for it to fall back out once that lift fades, because the underlying value never changed. That is temporary indexing, not a fix, and reading it as success leads people to keep mashing the button instead of addressing the page.
So stop treating the request as a remedy and use it for what it does: faster crawling of genuinely new or improved pages. If a page will not stay indexed, the answer is not another request, it is fixing the page, then requesting a crawl so Google sees the improved version sooner. The button speeds up the looking, not the deciding.