Record large-device retry diagnostics · 4e072b7

+19 -5

HealthProbe/Doc/04-project/Import-Optimization-Log.md

@@ -773,6 +773,19 @@ The likely bottleneck is per-row SQLite work:
   evidence for the large-device partial run or for commit `0d61338`. A matching
   Diagnostics text copy is still needed before drawing performance conclusions
   from the large-device partial run.
+- The matching Diagnostics report for that retry was captured later. Build
+  fingerprint was `1.0(1)-1780695759-92064`, operation result was
+  `partial_success`, `previousSnapshotID` was `none`, `isChainStart` was `true`,
+  but capture modes were `unchangedDelta=121`, `delta=4`, `initialImport=0`,
+  `unavailable=2`. This confirms the run was a retry using previously persisted
+  per-type capture state, not a clean first import rewrite of `8.4M` rows. The
+  performance result is still important for large-database retry/incremental
+  behavior: wall clock `11.2s`, fetch `6.0s`, processing `0.0s`, insert `1.0s`,
+  finalize `1.9s`, residual `1.5s`, with `793` delta events. The two failed
+  Blood Pressure metrics failed before record import because earliest/latest
+  queries returned HealthKit `com.apple.healthkit` error code `5`:
+  `Authorization not determined`. Treat this as a permission/device-state issue,
+  not an import engine regression.
 
 ## Open Issues / Observations
 
@@ -845,11 +858,12 @@ Prioritize experiments in this order:
    post-import responsiveness. Expected success: record archive finalize is
    zero/negligible, details still load from SQLite, and first-import memory /
    post-save behavior does not regress.
-11. Capture the matching Diagnostics text for the large-database retry after
-   first import that showed `8,411,713` records in `11.2s`. Confirm the build
-   identity, `CaptureModes`, and failed Blood Pressure API call status/details.
-   Interpret Blood Pressure failures in the context of this device's known
-   frozen / partially syncing Health database state.
+11. On the large-database device, resolve or deliberately document the two
+   Blood Pressure permissions. The matching report shows HealthKit
+   `Authorization not determined` for systolic/diastolic earliest/latest
+   queries, so the next app-side check is whether `Request Health Access`
+   re-prompts for those types or whether iOS Health settings expose them for
+   HealthProbe.
 12. Investigate full-profile empty anchored-query cost for zero-count types.
    Compare slow empty types across reports before changing behavior; any skip or
    lower-frequency strategy must preserve the promise that full authorized


	@@ -773,6 +773,19 @@ The likely bottleneck is per-row SQLite work:
773	773	evidence for the large-device partial run or for commit `0d61338`. A matching
774	774	Diagnostics text copy is still needed before drawing performance conclusions
775	775	from the large-device partial run.
	776	+- The matching Diagnostics report for that retry was captured later. Build
	777	+ fingerprint was `1.0(1)-1780695759-92064`, operation result was
	778	+ `partial_success`, `previousSnapshotID` was `none`, `isChainStart` was `true`,
	779	+ but capture modes were `unchangedDelta=121`, `delta=4`, `initialImport=0`,
	780	+ `unavailable=2`. This confirms the run was a retry using previously persisted
	781	+ per-type capture state, not a clean first import rewrite of `8.4M` rows. The
	782	+ performance result is still important for large-database retry/incremental
	783	+ behavior: wall clock `11.2s`, fetch `6.0s`, processing `0.0s`, insert `1.0s`,
	784	+ finalize `1.9s`, residual `1.5s`, with `793` delta events. The two failed
	785	+ Blood Pressure metrics failed before record import because earliest/latest
	786	+ queries returned HealthKit `com.apple.healthkit` error code `5`:
	787	+ `Authorization not determined`. Treat this as a permission/device-state issue,
	788	+ not an import engine regression.
776	789
777	790	## Open Issues / Observations
778	791
	@@ -845,11 +858,12 @@ Prioritize experiments in this order:
845	858	post-import responsiveness. Expected success: record archive finalize is
846	859	zero/negligible, details still load from SQLite, and first-import memory /
847	860	post-save behavior does not regress.
848		-11. Capture the matching Diagnostics text for the large-database retry after
849		- first import that showed `8,411,713` records in `11.2s`. Confirm the build
850		- identity, `CaptureModes`, and failed Blood Pressure API call status/details.
851		- Interpret Blood Pressure failures in the context of this device's known
852		- frozen / partially syncing Health database state.
	861	+11. On the large-database device, resolve or deliberately document the two
	862	+ Blood Pressure permissions. The matching report shows HealthKit
	863	+ `Authorization not determined` for systolic/diastolic earliest/latest
	864	+ queries, so the next app-side check is whether `Request Health Access`
	865	+ re-prompts for those types or whether iOS Health settings expose them for
	866	+ HealthProbe.
853	867	12. Investigate full-profile empty anchored-query cost for zero-count types.
854	868	Compare slow empty types across reports before changing behavior; any skip or
855	869	lower-frequency strategy must preserve the promise that full authorized