On 2026-07-22 a player filed six issues in a single batch — the first real report after v2.0.0. Not one long complaint thread. Six items, triaged, with a screenshot. Four are fixed and live. One shipped half. One is still open.
The most useful thing in the batch turned out to be the one I got wrong first.
The lever I was about to pull
The worst item was debris. Their screenshot showed "The station is passing through a debris cloud" firing about ten times in one short window, and by the time they wrote it up, the walls and windows were gone.
I had a diagnosis ready before I finished reading: SpaceDustMinor carries weight: 44 in the meteor table, which is enormous next to its siblings. Drop the weight, drop the spam. It is a one-number change and it is wrong.
SpaceDustMinor is announcement: null. It does not print that line at all — the ten announcements in the screenshot were SpaceDustMajor, which is the louder event with 8–12 meteors per wave instead of 3–5.
Below fifteen players, dust is the only eligible meteor event. Every sibling requires minimumPlayers of 15 or 25; both dust events sit at 0. So at low population the weights are not deciding whether dust fires. They are only deciding Minor versus Major.
Which means cutting Minor's weight shifts selection toward Major. The fix I was about to ship would have made a low-population shift louder and more destructive, and I would have shipped it believing I had helped.
What was actually driving it
The real lever was a cooldown that had been turned almost off. The MeteorSwarm parent sets reoccurrenceDelay: 1 against an engine default of 30. On top of that, SOLREIGN's preset drives the meteor table with a ramping scheduler that reaches an event every 20–60 seconds at full pressure. Dust then re-fires as fast as that one-minute gate allows, forever.
So: SpaceDustMinor 1 → 8. SpaceDustMajor 1 → 15, which hard-caps its announcements at two per half hour. And SpaceDustMajor's wave size 8–12 → 4–6, because a single firing was stripping hulls on its own.
After the deploy I read reoccurrenceDelay: 8 off the running server's filesystem rather than trusting the repo. Worth doing: my first read of that file looked like 1, because the value quoted inside my own explanatory comment sat above the real one.
The three that were small
Captain and Head of Personnel dropped from ten hours to two. Both roles gate on a flat OverallPlaytimeRequirement, and ten hours is roughly ten full shifts on a server where this player was often the only person online. At that population it reads as a grind wall rather than a trust gate. Two hours still stops a brand-new account walking into Captain on day one, and it is clearable in an evening. It goes back up if concurrent population ever makes Captain a contested role.
The artifact resolve cue is quieter. The "artifact stopped pulsing" sound carried a +3f volume boost while its sibling failure sound carried no boost at all, which made it the loudest thing in xenoarchaeology — in a room where artifacts unlock over and over. It is -2f now. Still reads as a resolution, no longer dominates the room.
The two toy swords left the Salvage vendor. They are deliberate easter eggs, a Minecraft reference and a Terraria one, and the player did exactly what the vendor invited — tested whether they mined ore, and reported them as broken when they did nothing. That is a placement bug, not a content bug. A vendor full of working salvage tools implies working salvage tools, so the wink read as breakage. They moved to the contraband inventory, where the plushie carp already establishes novelty, not utility. The items themselves are unchanged. The joke should be found, not sold.
The one that shipped half
The player said the station AI was repetitive and slightly irritating. The root cause was the opposite of what I had assumed.
PROVIDENCE is not a timer-driven voice commenting on nothing. Eleven of its twelve categories are event-driven — shift start, audits, raids, the zoo breach, death. Only idle musings run on a clock, and that clock had been deliberately made population-scaled so the voice would "lean in as the station empties": 20–40 minutes normally, tightening to 8–15 minutes at low population.
Tripling the rate on an eight-line pool does not read as company. It reads as a stuck loop, and low population is exactly where this player was. So it is reversed — low population now widens to 30–60 minutes. The emptier the station, the more PROVIDENCE holds its tongue. Silence is what makes the next line land. The idle pool also went from 8 lines to 18.
There was already a test pinning the old cadence in place, and it failed the moment I made this change — which is what it was there for. I pointed it at the new direction instead of deleting it, so if anyone later speeds low-population chatter back up, it trips the same wire.
This one is not finished, and I want it on the page. A separate Human Resources directive ticker — the "Productivity Mandate check-in" lines that were actually in the player's screenshot — still runs on its own five-minute schedule and was not touched by any of this. If you play a round tonight you will still hear it. Calling this item fixed would be a claim you could disprove in five minutes, so I am not making it.
The one still open
Restoring station power in the first ten minutes is still a chore. At a population of one to three, every round starts with breaking into Engineering to light the generators before the station works. The player proposed the standard fix themselves: map in one of those debug generators and give the station infinite power.
There are three drafted options — a pre-charged buffer on the primary SMES units, a limited corporate standby generator sized for critical loads only, or removing the break-in entirely by opening the Engineering path and pre-placing a fuelled portable generator. My recommendation is the first, then the third if breaking in is still the complaint after the charge is tuned.
What is missing is the decision, not the options. This one is pacing rather than a number, and whichever way it goes changes what Engineering is for on a quiet night. Every option moves the chore somewhere rather than deleting it, and shipping the wrong one means unshipping it later in front of the same people who filed the report.
Why the open ones are on the page
The response to this batch went out in character — a memorandum from PROVIDENCE addressed to the crew, naming the reporter item by item, including the two it could not claim. It closed by noting this was their second useful report; the first was the green-flash shader bug that became dev blog number two.
The format is not decoration. Four fixes with nothing left open is a list of claims, and a changelog that only lists wins is not evidence of anything.
Six in, four out, one halfway, one still cooking.