DSP-1236 had been waiting for six days, and the clock icon beside it had not moved once.
Tuesday, January 11, 05:45. The ticket form showed the same grey line it had shown since the fifth: Awaiting approver (Display). Abel had built the alarm group in an afternoon. The scheduler agent had already tagged PL BKR4 as a called point on Hall 4's interruptible schedule. The bid card was waiting to read a confirmation it could trust. The only thing missing was a person with the right role, and the queue had nobody in that role.
She opened the on-call roster instead. The display engineer who had approved DSP-1228 at 01:14 on a Tuesday night between Christmas and New Year's, the one who had made the coast feeders' yellow go away by moving the whole system's intermediate-state timer to two seconds, was on call again this week. His name was on the roster now, so she finally knew it. R. Villarreal.
She paged him. She did not explain the alarm group in the page. She wrote: DSP-1236, one breaker, test record attached, need approval before the 07:00 interval if possible.
He called back at 06:10. He sounded like he was lying down. Behind his voice a fan was running, the kind people put on a nightstand.
"I can approve a group," he said. "I'm not approving a group without a test. Last spring somebody moved points into a custom group at Bloomington and the whole station went silent on the pane for a shift. Nobody knew until a relay tech walked in."
"Fair," she said. "I have a test."
She attached Abel's notes from Monday while he waited. PL BKR4 had been forced to 00 for 1.4 seconds and the row had gone amber. Then it had been forced to 11 and the row had gone to BAD POINT CALL FIELD. Both events had the RTU stamp in UTC, and next to each one she had written the offset by hand, UTC minus 6, because the pane's own tag still said CPT and was wrong.
She added one line he hadn't asked for.
DSP-1228 stays at 2 s for the coast. This does not touch the pane timer.
There was a pause on the line, the kind that meant he was reading something he had signed.
"Okay," he said. "Yeah. Okay."
He approved at 06:58. At 07:30 Abel loaded the group, and she put PL BKR4 on the display herself and checked it the slow way: one row, group DR-CALLED, 0 seconds on 11, 1 second on 00. Every other breaker in the Coastal Texas Allocation Belt stayed on the system timer.
She could have added more. Hall 2 had two called breakers, and Hall 6 had a tie that the scheduler leaned on whenever the QSE sent a congestion instruction. Abel had asked, in his two-line way, more? She looked at the desk note on the hook wall, eleven numbered lines in green ink and pencil, and went down them with her finger. For two weeks the note had been saying the same thing in different words: if you can't name it and say why, don't trust it.
She could name one breaker and say why. She wrote under line 10, in pencil:
DSP-1236 live Jan 11 07:30. PL BKR4 only. Add by name, not by feeder.
Nothing was called overnight. That was the first thing she checked on Wednesday at 06:10. The bid card's event list for Hall 4 was empty, the DR-CALLED row had stayed grey all night, and there was no breaker to credit.
Trey had done the arithmetic anyway.
The call sheet was on her keyboard, blank side up, and on it he had redone Tuesday, January 4, the day that had started all of this: the 1.8 seconds of 11 that hadn't alarmed, and the meter bin that had paid the breaker for a fall the scheduler had already started. He had taken it the way she had finally taken it herself, on the 07:10 to 07:15 settlement bin and not the one-minute EMS trend, because the bid card got paid on settlement bins and nothing else.
Stair (sched clock, time-weighted): 4.6 to 7.7 MW Breaker: 16.2 to 19.3 MW TH practice, not a keep.
It matched hers to the decimal. She hadn't shown him her sheet. He must have read it off line 11, which now said write the range, and worked backward from Saucedo's checkpoint log the way she had.
He had done one thing she hadn't. He had written the SOE times twice.
13:12:03.4 UTC = 07:12:03.4 CST (-6, Jan)
And he had drawn a box around the -6, pressing hard enough that the pen had dented the sheet.
The SOE pane above her desk still tagged every stamp CPT. Saucedo's tag fix had not been in Monday's build, and it was not in last night's either. The report drafter, the agent that wrote the morning summary from the pane and signed it auto, copied the tag exactly. So every morning a machine wrote a confident line that named the wrong clock, and every morning a person at this desk undid it in their head.
She thought about texting him. She had unmuted his thread a week ago and it still felt new. Nice would be a word he'd have to decide what to do with. She didn't text.
She took the sheet to the hook wall and taped it under line 11 with two pieces of blue tape, so the next lead who read write the range would see what it looked like in somebody else's handwriting. In the corner she wrote checked MC 1/12. Then she sat down and added a line to the morning report, under the drafter's paragraph and outside it, so nobody would mistake it for the drafter's:
SOE tag still CPT on UTC stamps. Night desk converting by hand. Fix not in 1/11 build.
On Thursday she stopped writing notes about the tag.
05:50, January 13. The 1/12 build was live. The SOE pane said CPT. That made two releases since she had first asked, and her asking had taken the form of a sentence at the end of a longer email, which was the kind of request that goes into somebody's head and stays there.
The scheduler team's queue had a change request form with a severity field and an impact box. The form wanted a title under eighty characters. She wrote:
DSP-1241 · SOE pane labels RTU stamps CPT; stamps are UTC. Relabel to UTC or convert, do not do both.
She attached three things she already had instead of describing them. The first was Monday's test-position stamp, 11:22:41.306, beside a screenshot of the EMS banner reading 05:22. The second was her own wrong first offset from the tenth, with the correction next to it and the tick and the date, because if she was going to say the tag made people do arithmetic badly at five in the morning, she should show that it had made her do it badly too. The third was Trey's practice sheet with the boxed -6, photographed on the wall with the tape showing.
The impact box was where these requests usually went soft. People wrote confusing for operators and the queue read it as cosmetic. She wrote the part that made it a dispatch problem.
In March this becomes -5. Every lead converting by hand has to remember the date. Fix before Mar 12 2028.
She copied Saucedo. She asked him for one answer by Friday, which option and which build. She did not ask him for an apology, because an apology would take him longer to write than the answer.
Then she went to the hook wall and changed line 8 in pen. It had said subtract 6 (CST, Nov to Mar), subtract 5 (CDT, Mar to Nov) since the tenth, which was true and also asked a tired person to know which half of the year it was. She struck the pencil and wrote:
Subtract six until Mar 12, then five. Check the tag first. If it says UTC, subtract nothing more.
Saucedo answered at 05:12 on Friday, inside the deadline.
It was one paragraph. Option one: relabel. The 1/19 build would change the SOE tag from CPT to UTC and touch nothing else. The display team would not put daylight-saving logic in a pane that fed the event archive. Archive stamps had to be UTC end to end, because every protection engineer in three states pulled from it and none of them wanted a pane deciding what time an event happened. Conversion would be a separate request. It was not scheduled.
She read it twice at 05:35, standing, with her coat still on. Behind her the drafter had already posted Friday's summary to the morning channel, every SOE line in it tagged CPT, signed auto.
It was the right answer for the archive. It was also the answer that left the desk exactly where it was, subtracting by hand, with the number changing overnight in March. The relabel would make the stamp honest. It wouldn't make anyone faster.
She replied in the DSP-1241 thread, not in the canteen and not on the phone, so the reply would sit next to the request.
Relabel accepted for 1/19. It makes the stamp honest. It does not make the desk faster: every lead still subtracts by hand, and on Mar 12 the number changes overnight. Please open the conversion as its own CR now with a due date of Mar 1, so it is in somebody's queue before the clocks move, not after.
Then she asked for the one thing she actually cared about.
The 1/19 release note should say the old CPT tag was wrong on every stamp since the pane went live. Last year's event logs are UTC too. Nobody should read them as local.
She had seen what happened to wrong labels once they were in an archive. The drafter read old summaries to learn its phrasing. The bid agent read old event lists to calibrate its shed credit. If nobody wrote down that the tag had been lying, both machines would keep learning from it.
At the wall she rewrote line 8 for the third time, in pen. She did not erase the second version. She put one straight line through it, so Trey could still read what it had said and see what had changed.
Subtract six until Mar 12, then five. Check the tag first. If it says UTC, subtract nothing more.
Tag says UTC from 1/19. Subtract 6 until Mar 12, then 5. Before 1/19 the tag lied: those stamps are UTC too.
Saucedo replied at 06:20 with one line: CR-1250, due 3/1, release note updated. She printed it and taped it beside line 8, because a promise in a queue was easier to keep track of when it was also on the wall.
Saturday was her day off. The phone went at 06:58 while the kettle was still ticking.
"It's Trey." He was talking fast, then made himself slow down. "Sorry. It's a Saturday."
The cold front had come through at 03:00. She had heard it in her sleep, the window rattling in its frame and then the wind dropping all at once. By five the ERCOT price in the coastal zone had spiked hard enough for the QSE's agent to call interruptible load, and it had called Hall 4 for the 06:45 interval. The scheduler had started checkpointing at 06:31. PL BKR4 had opened on the call at 06:44:57.
"Under DR-CALLED the row went amber," he said. "00 for 1.3 seconds before it read open. Not 11. So it's not your exception."
"No."
"I called anyway." She heard him breathe in. "Your line says 11 over two seconds on a called point. This was 00. But a breaker that travels in fifty milliseconds doesn't sit in transit for a second and a third. Nobody else is watching that point on a Saturday. The bid card took it. It says confirmed."
She stood at the counter in her socks. The kettle clicked off. On the window ledge her phone's weather widget said 31 degrees in San Antonio, which meant colder on the coast.
He was right, and her line was wrong. She had written the exception the week before about the failure she had already seen: a bad point, 11, the breaker reporting both open and closed at once. She hadn't written it for the failure she hadn't seen yet, which was a breaker that took over a second to get from closed to open while the display waited, the bid card waited, and then both of them decided it was fine. A 00 that lasted that long wasn't a display problem. It was either the mechanism, a slow trip coil or a stiff linkage that got slower every time it operated until one day it didn't open at all, or the auxiliary switch that told the RTU where the contacts were, lagging behind contacts that had parted on time. The SOE couldn't tell those apart. The relay could.
She decided out loud, so he could write it down while she said it.
"You were right to call. The exception covers what I wrote, and what I wrote was too narrow. On a called point, any 00 over one second is a call to me and a work order to Abel. Tell him to pull the protection relay's event report for 06:44:57. The moment the current goes to zero is when it actually cleared. If that's on time, it's the aux switch. If it's late, it's the mechanism. And put the relay-house temperature on the work order. It was below freezing out there at three."
"Okay." She could hear the pen.
"Don't credit it yet. Do the bin the way you did on the twelfth. 06:45 to 06:50 settlement, the first full bin after the open. Subtract the stair from the scheduler log. Write the range."
"Range. Yes."
"And write the SOE stamps both ways. Both of them, where the 00 starts and where the 01 lands. The 1.3 seconds is all Abel has until he gets the relay record. The tag still says CPT until Wednesday."
He read it back to her. "00 at 12:44:55.9 UTC, 01 at 12:44:57.2 UTC. 06:44:55.9 to 06:44:57.2 CST. Minus six. 00, 1.3 seconds."
She heard the pen go around the six twice. He was boxing it.
"The bid card's going to want a number by eight," he said.
"Then it gets a range at eight, and point manual if it won't take one. That's what line 11 is for."
When he'd hung up she poured the water, which had gone a little cool, and stood with the cup at the window. The street was white with frost on the side the sun hadn't reached. Somewhere on the coast, PL BKR4 was open in a cold relay house, waiting for the scheduler to give Hall 4 its load back, and its mechanism had taken 1.3 seconds to do something that should take a twentieth of that. It had alarmed only because she had spent six days getting one person to approve one row. It had been noticed only because Trey hadn't read her rule as the limit of what he was allowed to worry about.
On Monday she would go to the hook wall before she took her coat off. She would strike over 2 s from line 11, because two seconds was the system timer that had let January 4 through in silence, and the DR-CALLED row she had spent six days getting approved alarmed on any 11 at all. The paper should match the instrument, the same as line 8. Then she would add the clause in pen, so the line read:
Call if any 11, or 00 over 1 s, on a called point. Relay record says which part.
Then she would take Trey's Saturday call sheet, initial it, date it, and tape it beside the clause. That way the next person to read the line would know it had changed because of his call, and that she had been wrong about where the edge was.
Every other line on that wall was in her handwriting, and that line would be the first one that wasn't only hers. She drank the cool tea and wrote relay record 06:44:57, coil or aux? Abel on the back of an envelope, so she'd remember it before she remembered anything else on Monday.