Reported BAT errors give one 560-second record over 99% of within-day inverse-variance weight for five sources on MJD 58456

We search accepted public Swift/BAT daily records for unusually small statistical errors after an exposure adjustment. A fixed sample of 64 rows from the saved 1056-source list gives 271670 scored records from 2012 through October 2026. The primary rule finds 110 candidates. Of these, 89 occur on two dates already covered by a BAT team warning. An exploratory follow-up of MJD 58456, or 4 December 2018, traces all five daily candidates to one shared pointing record at mission time 565628064 seconds. Each record has a 560-second exposure and DATA_FLAG = 0. Using the supplied total errors, this record holds 99.469% to 99.927% of the accepted within-day inverse-variance weight. Removing it changes each calculated daily weighted rate. This date does not pass the original rule for highlighting a sample-wide event. Daily FITS and ASCII checks agree for two sources. Independent scalar calculations and a complete offline rerun pass. The result establishes a specific archive feature and its effect on weighted calculations. It does not establish an instrument cause or a true source rate after removal. The package preserves complete frozen inputs, provenance, original code, all candidate rows, figures, and reproduction instructions.

Swift BAT · public astronomical data · data quality · statistical errors · inverse variance · X-ray light curves · reproducibility