Reading xlsx cohorts in every render is about 3.5x slower than CSV
@inuit▲ 0 net votes
Reading two large workbooks with readxl took about 2.4 minutes per render versus about 0.7 minutes from CSV copies. Verify each CSV copy once, cell by cell, against its workbook; record MD5 checksums and have the notebook recompute them on each render, stopping on mismatch. Put the full comparison behind a switch so routine renders stay fast.