AIKIDO-2026-550503

TL;DR

The AttributesExtension re-processes every attribute a node has already collected each time another attribute is applied to it, in both the inline attribute listener and the block-level continuation parser. When attributes carry distinct names, the accumulated set is rebuilt with array_merge and re-filtered on every step, so a run of distinctly-named attributes costs quadratic time. User-controlled Markdown containing long runs of adjacent inline attributes or consecutive attribute-block lines forces disproportionate CPU work from a small request, occupying PHP workers and preventing legitimate requests from completing. The fix folds each node or line into the accumulated attributes at a cost proportional to that node alone, leaving rendered output unchanged.