Undecided. This issue records the question and what's known so far.
Background
#48 changes how a fault still listed at the last poll ends. It used to end at ESB's estimate even when that estimate had already passed; it now runs to the collection horizon. The notes had promised the elapsed time as a lower bound, and the model didn't deliver it.
Planned works were left alone. A planned job still listed at the horizon keeps ending at its scheduled end (end_src "estimated"), as before. The reason given in notes/grading.md § A live fault runs to the horizon is the one notes/design-alignment.md already uses for the row wording: "a listing is not an observed outage".
The case that raised it
At the 24 September horizon, Kilcoole (Wicklow) is planned work scheduled to end 10 September 16:00 UTC, and it is still listed. Run to the horizon, it would be about 332 hours long. Wicklow's September day bar would mark two weeks as planned, where it now marks the scheduled days only. Grades are unaffected either way, because planned works are excluded from them. Planned customer-minutes would grow.
The question
Is a planned job still on ESB's list past its end date:
- (a) a stale listing (ESB forgot to take it down), so the schedule is the better end, which is the current behaviour; or
- (b) a job that overran, so the listing is the better evidence, and it should run to the horizon like a fault?
Nothing in the record tells the two apart. It would help to measure how often a planned job outlives its schedule, and by how much, across the corpus: a few minutes suggests an overrun, days suggest a stale listing. A cap could come out of that (for example, run to the horizon for up to N hours past the schedule).
Undecided. This issue records the question and what's known so far.
Background
#48 changes how a fault still listed at the last poll ends. It used to end at ESB's estimate even when that estimate had already passed; it now runs to the collection horizon. The notes had promised the elapsed time as a lower bound, and the model didn't deliver it.
Planned works were left alone. A planned job still listed at the horizon keeps ending at its scheduled end (
end_src"estimated"), as before. The reason given innotes/grading.md§ A live fault runs to the horizon is the onenotes/design-alignment.mdalready uses for the row wording: "a listing is not an observed outage".The case that raised it
At the 24 September horizon, Kilcoole (Wicklow) is planned work scheduled to end 10 September 16:00 UTC, and it is still listed. Run to the horizon, it would be about 332 hours long. Wicklow's September day bar would mark two weeks as planned, where it now marks the scheduled days only. Grades are unaffected either way, because planned works are excluded from them. Planned customer-minutes would grow.
The question
Is a planned job still on ESB's list past its end date:
Nothing in the record tells the two apart. It would help to measure how often a planned job outlives its schedule, and by how much, across the corpus: a few minutes suggests an overrun, days suggest a stale listing. A cap could come out of that (for example, run to the horizon for up to N hours past the schedule).