DEV Community has published a developer's checklist for sitemap errors in Google Search Console. The two messages in question are "Couldn't fetch" and "Sitemap could not be read". The article says the first can appear simply because the status has not caught up, so a new sitemap should be given a day before any debugging.
If the error stays for a day or two, the article advises requesting the sitemap with curl, a command-line tool, using Googlebot's user agent from a machine outside your own network. A browser can mislead because it carries cookies and may have passed a bot challenge. You should see a final status of 200, ideally with no redirects, an XML content type, and a body that starts with <?xml and contains <urlset or <sitemapindex.
The article lists eight usual causes. These are a firewall or bot protection blocking Googlebot, robots.txt rules that block the file, redirect problems, a page that is not really XML, and malformed XML such as unescaped ampersands. The rest are gzip confusion, a file over 50,000 URLs or 50 MB uncompressed, and a sitemap that is slow to generate and times out. For firewalls, it recommends allowing verified Googlebot traffic rather than trusting the user agent string, which anyone can fake.
Once the file is readable, the article says to check the URLs inside it. Each should be the canonical, indexable version that returns 200. It also says sitemap index files need every child sitemap to pass the same checks, including matching host and protocol. After fixing problems, resubmit the sitemap. If the status does not update, the URL Inspection tool's live test shows what Googlebot received. The article is from a vendor of a monitoring tool, which it mentions as a way to catch such problems after deploys.
Why it matters
A sitemap that Google cannot read slows down how it discovers your pages. The article points out that the cause is often outside the sitemap file itself, such as a firewall rule or a framework upgrade, and that Search Console can take days to report it. A site can be affected for some time before anyone notices.
Most of the checks are quick and need no special tools. Testing from outside your network as a crawler is the key step, because browser tests can hide the problem.
Source: DEV Community. This brief summarises their reporting; read the original for the full detail.