• GeneFox no longer quits without warning on Mac. A background task inside the cloud-sync library could fail writing to its own cache and take the whole app down, with no chance for GeneFox to intervene. That cache is gone; nothing depended on it. • A saved report that refused to reopen now opens. A report holding a figure captured at a high colour depth — usually a heatmap — saved correctly, then called itself corrupt on every later load. It was neither corrupt nor too large. Reports already stuck this way open now. • A pan-cancer figure exports whole instead of being quietly cropped. Exporting all 54 cohorts at a single-column width cut off the lower cohorts and baked the provenance text into the image, without saying so. • A project list no longer empties because one project came from a newer version. A single unreadable project stopped the whole list loading. It is now skipped and the rest open. Saved analysis recipes had the same flaw and the same fix. • A saved view you opened once no longer reopens itself days later. Some corrections change figures or numbers you may already have exported. If you relied on any of these, it is worth regenerating the figure: • Compare Cohorts no longer reports a p-value across arms that share patients. With adjacent normal included, a cohort contributes both its tumour arm and its own adjacent-normal arm — the same people. Those groups are not independent, and a p-value obtained that way is not one to report. • Box plots no longer drop their most extreme outliers. A box drew at most 100 outliers and took the first 100 of an ascending list, so a box with many low outliers showed none of its high ones, and the axis shrank to match. • A heatmap gene row containing an infinity now scales correctly. Such a row was drawn flat and mid-toned across every sample — the same picture as a gene that genuinely does not vary — rather than as a visible gap. • The gene-correlates figure's threshold lines now sit where its caption says they do. • Small p-values are printed the same way everywhere. No p-value changed; only how a small one is displayed.