Record large-device partial report mismatch · c089a7d

+22 -9

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

@@ -759,6 +759,15 @@ The likely bottleneck is per-row SQLite work:
   blob on fresh first imports only adds CPU, memory, and SwiftData payload size.
   Existing optional fields remain for old rows and compatibility, but new
   SQLite-backed captures should report no compact archive finalize cost.
+- A 2026-06-06 screenshot from the large-database device showed a fast partial
+  full-profile result: `127` metrics, `8,411,713` records fetched, `11.2s`
+  duration, and `2` failed metrics (`Blood Pressure Diastolic` and
+  `Blood Pressure Systolic`). The attached copied diagnostic text did not match
+  that screenshot; it was the older `1.0(1)-1780689890-92064` repeated-capture
+  report with `2,625,217` records and `5.8s` duration. Do not use that pasted
+  text as 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.
 
 ## Open Issues / Observations
 
@@ -831,21 +840,25 @@ 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. Investigate full-profile empty anchored-query cost for zero-count types.
+11. Capture the matching Diagnostics text for the large-database partial run
+   that showed `8,411,713` records in `11.2s`. Confirm the build identity,
+   `CaptureModes`, failed Blood Pressure API call status/details, and whether
+   the run was a repeated delta, partial resumed import, or clean first import.
+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
    backup can notice newly appearing data.
-12. Run a non-chain-start/full-scan benchmark after skipping unchanged `verified` events and fast-pathing already-open visibility ranges. Compare `SummedInsertElapsed`, `Heart Rate insertElapsed`, `Steps insertElapsed`, and `Walking + Running Distance insertElapsed`.
-13. Reduce any remaining per-sample SQLite writes for unchanged existing samples during non-chain-start full scans.
-14. Profile whether index maintenance dominates first-import insert cost.
-15. Consider a guarded bulk-import mode for first observations:
+13. Run a non-chain-start/full-scan benchmark after skipping unchanged `verified` events and fast-pathing already-open visibility ranges. Compare `SummedInsertElapsed`, `Heart Rate insertElapsed`, `Steps insertElapsed`, and `Walking + Running Distance insertElapsed`.
+14. Reduce any remaining per-sample SQLite writes for unchanged existing samples during non-chain-start full scans.
+15. Profile whether index maintenance dominates first-import insert cost.
+16. Consider a guarded bulk-import mode for first observations:
    - keep archive semantics unchanged;
    - only relax work that can be safely reconstructed or validated;
    - re-enable normal idempotent paths for incremental observations.
-16. Run a fresh first-import benchmark after the unused-index removal and compare `SummedInsertElapsed`, `Heart Rate insertElapsed`, and `Active Energy insertElapsed`.
-17. Investigate whether first-import-only deferred index creation or temporary staging tables can reduce `samples` / `sample_versions` / `sample_observation_events` write cost without weakening final archive integrity.
-18. Revisit adaptive page sizes only after SQLite write-path costs are reduced.
-19. Revisit background / scheduled collection once initial import can finish reliably and post-import UI recovery is bounded.
+17. Run a fresh first-import benchmark after the unused-index removal and compare `SummedInsertElapsed`, `Heart Rate insertElapsed`, and `Active Energy insertElapsed`.
+18. Investigate whether first-import-only deferred index creation or temporary staging tables can reduce `samples` / `sample_versions` / `sample_observation_events` write cost without weakening final archive integrity.
+19. Revisit adaptive page sizes only after SQLite write-path costs are reduced.
+20. Revisit background / scheduled collection once initial import can finish reliably and post-import UI recovery is bounded.
 
 ## Verification Checklist For Each Optimization
 


	@@ -759,6 +759,15 @@ The likely bottleneck is per-row SQLite work:
759	759	blob on fresh first imports only adds CPU, memory, and SwiftData payload size.
760	760	Existing optional fields remain for old rows and compatibility, but new
761	761	SQLite-backed captures should report no compact archive finalize cost.
	762	+- A 2026-06-06 screenshot from the large-database device showed a fast partial
	763	+ full-profile result: `127` metrics, `8,411,713` records fetched, `11.2s`
	764	+ duration, and `2` failed metrics (`Blood Pressure Diastolic` and
	765	+ `Blood Pressure Systolic`). The attached copied diagnostic text did not match
	766	+ that screenshot; it was the older `1.0(1)-1780689890-92064` repeated-capture
	767	+ report with `2,625,217` records and `5.8s` duration. Do not use that pasted
	768	+ text as evidence for the large-device partial run or for commit `0d61338`.
	769	+ A matching Diagnostics text copy is still needed before drawing performance
	770	+ conclusions from the large-device partial run.
762	771
763	772	## Open Issues / Observations
764	773
	@@ -831,21 +840,25 @@ Prioritize experiments in this order:
831	840	post-import responsiveness. Expected success: record archive finalize is
832	841	zero/negligible, details still load from SQLite, and first-import memory /
833	842	post-save behavior does not regress.
834		-11. Investigate full-profile empty anchored-query cost for zero-count types.
	843	+11. Capture the matching Diagnostics text for the large-database partial run
	844	+ that showed `8,411,713` records in `11.2s`. Confirm the build identity,
	845	+ `CaptureModes`, failed Blood Pressure API call status/details, and whether
	846	+ the run was a repeated delta, partial resumed import, or clean first import.
	847	+12. Investigate full-profile empty anchored-query cost for zero-count types.
835	848	Compare slow empty types across reports before changing behavior; any skip or
836	849	lower-frequency strategy must preserve the promise that full authorized
837	850	backup can notice newly appearing data.
838		-12. Run a non-chain-start/full-scan benchmark after skipping unchanged `verified` events and fast-pathing already-open visibility ranges. Compare `SummedInsertElapsed`, `Heart Rate insertElapsed`, `Steps insertElapsed`, and `Walking + Running Distance insertElapsed`.
839		-13. Reduce any remaining per-sample SQLite writes for unchanged existing samples during non-chain-start full scans.
840		-14. Profile whether index maintenance dominates first-import insert cost.
841		-15. Consider a guarded bulk-import mode for first observations:
	851	+13. Run a non-chain-start/full-scan benchmark after skipping unchanged `verified` events and fast-pathing already-open visibility ranges. Compare `SummedInsertElapsed`, `Heart Rate insertElapsed`, `Steps insertElapsed`, and `Walking + Running Distance insertElapsed`.
	852	+14. Reduce any remaining per-sample SQLite writes for unchanged existing samples during non-chain-start full scans.
	853	+15. Profile whether index maintenance dominates first-import insert cost.
	854	+16. Consider a guarded bulk-import mode for first observations:
842	855	- keep archive semantics unchanged;
843	856	- only relax work that can be safely reconstructed or validated;
844	857	- re-enable normal idempotent paths for incremental observations.
845		-16. Run a fresh first-import benchmark after the unused-index removal and compare `SummedInsertElapsed`, `Heart Rate insertElapsed`, and `Active Energy insertElapsed`.
846		-17. Investigate whether first-import-only deferred index creation or temporary staging tables can reduce `samples` / `sample_versions` / `sample_observation_events` write cost without weakening final archive integrity.
847		-18. Revisit adaptive page sizes only after SQLite write-path costs are reduced.
848		-19. Revisit background / scheduled collection once initial import can finish reliably and post-import UI recovery is bounded.
	858	+17. Run a fresh first-import benchmark after the unused-index removal and compare `SummedInsertElapsed`, `Heart Rate insertElapsed`, and `Active Energy insertElapsed`.
	859	+18. Investigate whether first-import-only deferred index creation or temporary staging tables can reduce `samples` / `sample_versions` / `sample_observation_events` write cost without weakening final archive integrity.
	860	+19. Revisit adaptive page sizes only after SQLite write-path costs are reduced.
	861	+20. Revisit background / scheduled collection once initial import can finish reliably and post-import UI recovery is bounded.
849	862
850	863	## Verification Checklist For Each Optimization
851	864