Known bugs and error patterns. Agents update when found/fixed.
Root cause: PyMuPDF 2x render scale produced images too small for zxing-cpp to decode barcodes on these files. The .ai file rendered at 1099x415 vs the .jpg source at 2290x866.
Fix: Bumped render scale from 2x to 3x in loader.py. All 4 files now decode correctly. Tests updated — xfail markers removed, assertions upgraded to verify extracted values.