Say when a skip is probing, cross damage in doubling steps, and stop …

[olavgg]

…repeating advice already taken

A real drive on --skip-slow looked hung for minutes: probe reads were in no latency window and the percentage only follows the scan, so a skip probing ahead was indistinguishable from a read that never came back -- and at a fixed gigabyte per step, crossing a drive damaged to the end is thirteen thousand probes, hours of it.

• While a skip probes, the drive's state is "probing" and the line under it says where ("skip: probing ahead at X for where the slow region ends"); so does a single-drive scan's status line. Probe reads count in the latency window, and the percentage follows the last point known slow. The checkpoint keeps the real position, since nothing probed is condemned until the skip finishes.
• Forward steps start at 1 GiB and double after each slow probe, up to 64 GiB: a couple of hundred probes to cross a 12.7 TiB drive. A healthy stretch shorter than the step between two slow probes is condemned with them.
• The stall warning told a run given '--retries 3 --recovery-time 300' to re-run with exactly that. It now names only the levers not in effect, adds --skip-slow, and stays quiet under --skip-slow.