investigate: camera FigCaptureSource errors and cold start exceeding 0.5s target

[jukasdrj]

Context

During a successful scan test (Feb 20, 2026), the console logs show several camera subsystem errors and performance target misses. The scan still completed successfully, but these warrant investigation.

Observed Issues

1. FigCaptureSource/FigXPCUtilities Errors

```
(Fig) signalled err=-12710 at <>:601 (x3, at startup)
signalled err=-17281 at <>:308 (x4+, during session)
Fig assert: "err == 0" at bail (FigCaptureSourceRemote.m:569) - (err=-17281)
Fig assert: "err == 0" at bail (FigCaptureSessionRemote.m:910) - (err=-12784)
```

These are internal AVFoundation/FigCapture errors. They may be:

• Simulator-only artifacts (not present on real device)
• Related to camera session configuration timing
• Benign warnings that don't affect functionality

2. Camera Cold Start: 1.996s (target: < 0.5s)

```
Camera session configured in 1.995s
Camera cold start: 1.996s (target: < 0.5s)
Camera cold start exceeded 0.5s target!
```

4x over target. May be simulator overhead — needs real device profiling.

3. Image Processing: 1.209s (target: < 0.5s)

```
Image processed in 1.209s (target: < 0.5s)
Image processing exceeded 0.5s target!
Preprocessing: 68ms, rotated: false, brightness adj: 0.200000
```

Preprocessing itself was fast (68ms) but total processing was 1.2s. The gap (~1.1s) may be JPEG compression or other pipeline steps.

4. QUIC/Network Warnings

```
quic_crypto_queue_append not enqueing more packets (max 5 reached) (x4)
nw_connection_copy_protocol_metadata_internal on unconnected nw_connection (x3)
```

These are iOS networking stack warnings during the upload. Likely benign but worth understanding.

5. numANECores: Unknown aneSubType

Logged once — may be simulator not having Apple Neural Engine info.

Investigation Steps

• Reproduce on real device — are Fig errors simulator-only?
• Profile camera cold start on device (Instruments → Time Profiler)
• Profile image processing pipeline — where is the 1.1s gap between preprocessing and total?
• Determine if QUIC warnings indicate a real connectivity issue or are cosmetic
• Check if ANE warning affects any ML processing paths

Priority

Low — scan completed successfully with all 7 books identified. These are performance/quality improvements, not functional bugs.

Labels

investigation, performance, epic-6