# Remediate HPT2 graph reconstruction performance

Status: complete
Type: campaign
Updated: 2026-08-27
Next Action: none
Campaign ID: hpt2-graph-reconstruction-performance-remediation
Campaign Number: 024
Outcome: Produce a new exact HPT2 candidate whose initial graph reconstruction preserves mirrored integrity and completes within 60 seconds on both controller variants at the full 142,560-record supported ceiling.
Primary Focus Areas: hpt-media-logging
Supporting Focus Areas: qualification-release
Depends On: none
Decision: none
Detour For: none
Return To: none
Completion Gate: The frozen predecessor remains unchanged; the mirror-safe fast path and explicit full-recovery claim decision are documented; focused checks and canonical Debug/Release builds pass; CID-linked Controller A and B evidence passes first at 56,160 records and then at 142,560 records with initial graph reconstruction no greater than 60 seconds on each controller; downstream wrap/reclaim, export, retail-card, ECO, and rollout work remains deferred.
Completion Evidence: work/evidence/hpt2-graph-reconstruction-remediation-target-gate-20260827.md
Completion Date: 2026-08-27
Completion Order: 11
Disposition: completed

## Request

The frozen predecessor measured 61.82-62.59 seconds for 56,160 records because
the graph projection read data and commit blocks from all three mirrors. Keep
that exact candidate and its narrowing evidence unchanged. Implement a
graph-only bounded route that validates a committed primary copy and falls back
to the existing full-quorum resolver for a complete window on any failure.
Do not change generic exact reads, export, diagnostics, scrub, write, reclaim,
or mirror-integrity semantics.

Formally decide whether full recovery is a supported bounded product/operator
surface. Freeze a new clean exact candidate after focused checks and canonical
Debug/Release builds. Capture same-run controller/CID evidence and test A and B
at 56,160 records first. Advance to 142,560 records only after both pass at no
more than 60 seconds. Keep wrap/reclaim, export, retail-card qualification,
ECO, and rollout deferred.

## App Server Preparation Contract

```json
{
  "campaign_id": "hpt2-graph-reconstruction-performance-remediation",
  "completion_evidence": "The frozen predecessor remains unchanged; the mirror-safe fast path and explicit full-recovery claim decision are documented; focused checks and canonical Debug/Release builds pass; CID-linked Controller A and B evidence passes first at 56,160 records and then at 142,560 records with initial graph reconstruction no greater than 60 seconds on each controller; downstream wrap/reclaim, export, retail-card, ECO, and rollout work remains deferred.",
  "exact_resolution": "dispatch-time",
  "execution_shape": "single-bounded-camp",
  "inline_assets": "metadata-only",
  "objective": "Produce a new exact HPT2 candidate whose initial graph reconstruction preserves mirrored integrity and completes within 60 seconds on both controller variants at the full 142,560-record supported ceiling.",
  "schema_version": 1,
  "source_freshness": "required",
  "verification": "orchestrator-exactly-once"
}
```

## Completion Check

The frozen predecessor remains unchanged; the mirror-safe fast path and explicit full-recovery claim decision are documented; focused checks and canonical Debug/Release builds pass; CID-linked Controller A and B evidence passes first at 56,160 records and then at 142,560 records with initial graph reconstruction no greater than 60 seconds on each controller; downstream wrap/reclaim, export, retail-card, ECO, and rollout work remains deferred.
