RETURN CONDITIONS
Origin date unknown // Multiple versions detected // Reconciliation failed
No one agrees on when this document was first written.
Some copies claim it began as a checklist.
Others insist it was always a story told to new staff so they'd know what not to ask.
By the time anyone thought to formalise it, the rules were already behaving strangely.
Everything else keeps shifting.
At first, returns were treated as exceptions.
A subject exited continuity, then reappeared, and the system adjusted around them.
Early models assumed this would be rare.
It wasn't.
Patterns emerged. Some exits left no residue. Others left too much.
Witnesses remembered things they shouldn't. Objects persisted without explanation.
Administrative logs refused to close.
The language changed after that.
People stopped saying death and started saying exit event.
They stopped saying resurrection and started saying return.
This made the work easier to do.
A clean exit is supposed to be tidy.
Nothing carries across.
No witnesses remain attached.
No objects behave strangely.
No one has to improvise an explanation.
When a subject returns from a clean exit, they fit back into the timeline with minimal friction. They function. They comply. They do what the system expects of them.
They also tend to feel hollow.
Damaged exits are different.
Something breaks on the way out.
A memory tears. An object comes with them. A witness refuses to forget.
The system compensates, but the compensation leaves marks.
Damaged returns come back carrying those marks.
They loop.
They hesitate at the moment things are meant to conclude.
They remember events the timeline insists never happened.
The system does not consider this a failure.
It considers it a cost.
Unresolved exits are worse.
The subject may return, but the system keeps looking for them anyway.
Tasks remain open.
Forms stay active.
Doors disagree about whether they're locked.
In these cases, the return is treated as provisional, even if it lasts for years.
No one likes working on unresolved cases.
They accumulate quietly.
Over time, a set of practices emerged, though no one admits to authoring them.
If witnesses are present, you give them a cover story.
Nothing dramatic. Nothing intentional. Weather. Miscommunication. A clerical error.
If objects cross the boundary, you try to quarantine them. If that fails, you let them drift until people stop asking questions.
If a subject doesn't quite align with the current timeline, you don't reject them. You adjust the paperwork.
Some returns don't stabilise on their own.
In these cases, the system relies on anomalies it doesn't fully understand.
Carrier voices, for example.
Surfaces that can hold contradiction without resolving it. They don't fix anything, but things stop escalating around them.
Runners appear in the same files.
They don't complete tasks correctly.
They complete them enough.
A failure logged as success.
A delivery recorded without arrival.
A job left open until it no longer demands attention.
Officially, these are treated as noise.
Unofficially, they're tolerated because removing them makes things worse.
There was an attempt, once, to eliminate damaged returns altogether.
The proposal argued that clean exits were more efficient, more legible, less expensive to manage. It recommended stricter enforcement, tighter optimisation, fewer exceptions.
The proposal was approved.
It was also quietly abandoned.
The systems became brittle.
Small errors cascaded.
People began returning empty more often than intended.
The phrase started circulating again, this time with a different emphasis:
Most current copies of this document end without a conclusion.
Some trail off mid-sentence.
Some repeat the same warning twice.
Some repeat the same warning twice.
One version ends with a handwritten note that appears nowhere else:
No action was taken on this note.
The case is marked resolved.
No one remembers who first wrote it.
No one has managed to prove it
▒▒▒ DOCUMENT FRAGMENT ENDS ▒▒▒
ADDITIONAL DATA CORRUPTED
RECOVERY IMPOSSIBLE