AIKIDO-2026-785533

TL;DR

The GitHub Flavored Markdown Table extension's block start parser reads the entire accumulated paragraph buffer and uses strpos to search for a pipe character on every non-blank line while a paragraph is open. For a paragraph of M lines with no pipes, this searches roughly k lines of buffer on line k, giving O(M^2) total work that grows with the size of the submitted Markdown. Rendering a single large paragraph of lines with no pipes through GithubFlavoredMarkdownConverter, or any environment with the Table extension enabled, uses CPU time that scales quadratically with body size and can exhaust worker processes. The fix checks only the current line for a delimiter row before using the paragraph buffer, removing the per line rescan.