The municipal records office offered no drama.
It offered plastic chairs, slow terminals, and a clerk named Fatma who refused to let urgency become bad procedure.
“You brought five permit references,” she said. “We search five. No contractor-wide fishing.”
Musa agreed.
The first permit number returned Block 17, Mtoni.
The disputed lot had come from Upanga East.
“Mismatch,” Salma said.
“Maybe an internal job reference was entered into a permit field,” Fatma replied. “Do not call it a forged permit from this screen.”
The second case had a correct address but inconsistent appeal status. The third was clean. Musa was pleased to see it.
Salma noticed.
“You are happy because a case is clean?”
“It shows us what normal looks like.”
They used the clean permit as a control example.
Fatma explained the city field definition. The permit field was supposed to contain an official identifier, not a generic contractor job code. If Kifutio’s system used the field differently, that was a design problem even before anyone decided whether a specific removal was lawful.
They reviewed appeal stamps supplied by Salma through tenant counsel. In several cases, current appeal status existed in one city process but was not reflected in contractor materials.
“City feeds can be delayed too,” Fatma warned.
That became an alternative explanation.
Musa refused management’s request to use the words *forged permits*.
“What we have is enough: address mismatch and stale appeal status.”
“Commercial wants something stronger,” Asha said by phone.
“Precision is stronger if a regulator has to defend it.”
Musa ordered certified extracts for three representative cases rather than copying the entire municipal record universe.
Then Salma produced a live situation for the next morning.
A Kifutio clearance crew was scheduled to arrive at an apartment block where an appeal had been filed that day.
The tenant counsel had consented to share the reference.
Musa contacted the city housing ombuds office.
His request was narrow: observe whether current appeal status reached the contractor checklist before removal and resale triggers.
SokoPili’s role would remain limited to marketplace control.
Musa would not direct an eviction.
He would not enter the apartment.
He would not pretend platform authority was city authority.
The ombuds office accepted the observation.
Fatma handed over the certified extracts and said, “Do not use my terminal as evidence of motive.”
“We won’t.”
The investigation had left the marketplace and entered external regulation through a small number of verifiable records.
Fatma’s clean control case became part of the ombuds request alongside the defective examples. Musa wanted the external review to know what a normal municipal chain looked like: correct permit, matching address, current status, consistent contractor reference. That made the defect cases measurable rather than merely suspicious. The request also separated custody. The city held permit records. Tenant counsel held appeal documents. SokoPili held listing history. No one institution needed to copy the others’ full files. The live observation planned for dawn would test the interface between those systems rather than turning them into one giant database.
The ombuds officer reviewing the request asked whether SokoPili wanted the city to validate every Kifutio job. Musa said no. The immediate question was whether the observed control gap could be reproduced and then fixed. Broad inspection could follow only if evidence justified it. That answer surprised Salma, who had expected the platform to seek maximum leverage. Musa preferred a narrow test because a successful narrow test could support a defensible remedy. The dawn observation would therefore focus on one disputed unit, one current appeal, and one contractor checklist.
Musa added the clean permit example to his memo so the ombuds team could compare a functioning workflow against the defective one instead of interpreting every difference as misconduct.
Tomorrow at dawn, the workflow itself would be tested before the door closed.