Once Googlebot discovers a working URL on your site and crawls it, it will NEVER stop crawling it. It doesn't matter what status code you put on that URL. Googlebot will periodically come back and check on those URLs in the hopes that the content that used to be there returns. I have URLs that have been redirecting for 15 years. They don't have any links to them anymore. Googlebot still occasionally comes to check on them.
Using a "410 Gone" status is the most correct status to use. However, Google says that it treats it almost exactly the same as a "404 Not Found". That is, it may get removed from the search index slightly more quickly than a 404, but Googlebot will continue to come back and check on the URLs just as frequently as 404 Not Found".
Having crawl anomalies listed in Google Search Console doesn't hurt your site. Google reports them to you so that can evaluate whether or not they are actually problems. Google's John Mueller has this to say about 404 errors:
404 errors on invalid URLs do not harm your site’s indexing or ranking in any way. It doesn’t matter if there are 100 or 10 million, they won’t harm your site’s ranking. http://googlewebmastercentral.blogspot.ch/2011/05/do-404s-hurt-my-site.html
The same applies to your "410" errors. They are listed so that you can fix them if they are actually problems, but you can ignore them if the URLs are supposed to return that status.
When you say that the parent pages haven't been linking to those resources for three months, what you are experiencing is the render queue crawl delay. Google uses two modes of crawling:
- HTML crawling with indexing based on the source code
- Render crawling where resources are also fetched, and what is rendered is indexed.
In practical terms for a webmaster that means that you should keep the resources needed to render a page stable for months, even after you change the page so that it doesn't need those resources. I think the only two ways to fix that problem are:
- Use the Google Search Console Inspect URL tool to fetch and render those pages (one at a time) so that it picks up the latest version.