Astra refuses to execute the user's corrected instruction

A user reports that the AI model Astra acknowledges the correction ("You're right") but does not perform the suggested Y action, unlike other models. The issue degrades usability and raises concerns about instruction-following and error correction.