[yotamon]

Why

Vercel is prerendering 204 static pages using one worker. The public blog has 36 Markdown posts (each with English/Hebrew content). getAllPosts() and getPostBySlug() reread and remark-render full Markdown during each call, including calls from the homepage, sitemap, blog listing, and each prerendered article's page+metadata.

What changed

• Memoize parsed published posts by slug only in production, reusing concurrent promises. Restrict the post cache to known published slugs to prevent unbounded growth from requests for invalid slugs.
• Memoize the sorted list of all posts in production so static generation does not reread all 36 files on every call.
• Preserve uncached reads during development to keep live Markdown editing responsive.
• Cover production identity/parallel behavior, development re-reading, and missing post results.

Safety

• The Markdown files ship with the deployment and cannot change in a running production instance without a new release.
• No changes to route behavior, locale content, metadata, SEO, rendering semantics, or runtime API.
• No Vercel preview and no manual deployment; compare static-generation timings on the next automatic production deployment after merging.