What a rebuild really asksThe reading of every sector of every remaining drive, in an array of four drives of 16TB each, means about 48TB of continuous reads from drives the same age as the one that failed. This is the plain risk.
The URE number, honestlyA consumer hard drive is rated to have one unrecoverable read error (URE) every 1014 bits (which equates to about one bad sector in every 12.5TB read); an enterprise-class hard drive is rated at 1015, or ten times better than consumer. That rating is a warranty floor, not a schedule, and most drives do continue to read cleanly beyond their ratings. The rating is based on fact; the 'doom' graphs are marketing tools.
What controllers do when it happensBehaviour varies: older controller systems terminate the rebuild when they encounter a single unrecoverable read error. Modern PERC and MegaRAID controllers puncture the affected stripe(s), allowing them to continue rebuilding. Linux mdadm will log those errors into its bad-block file. One unreadable sector should cost one stripe, not the array.
Why the second failure is usually mechanicalDrives from the same production batch, with the same hours of operation and the same heat, then a rebuild's marathon read. The survivor that was quietly marginal will likely break down during this marathon process. Gentle imaging, weakest regions last, is how we run that marathon instead.