# Wetech Multiple-target Route Fallback

<!-- auto-oncall-optimization-record
{
  "schemaVersion": "auto-oncall-optimization-record-v1",
  "optimizationId": "wetech-multiple-target-route-fallback-2026-09-05",
  "problems": [
    {
      "problemId": "wetech-multiple-target-route-terminal-error",
      "summary": "A Wetech Scenario Router response containing multiple targets failed strict validation and terminated analysis instead of returning a bounded human handoff.",
      "status": "candidate",
      "issueKeys": ["APID-29465"],
      "sourceRefs": ["https://tidb.atlassian.net/browse/APID-29465#problem:current-wetech-multiple-target-route-fallback"],
      "signalTextIncludesAny": ["diagnostic_wetech_route_target_count_invalid", "multiple Wetech targets", "multiple_cluster_targets"],
      "observedThrough": "2026-09-05",
      "effectiveAt": null,
      "verifiedAt": null,
      "attemptedAction": "Convert only the strict Wetech multiple-target route validation error into the existing bounded human-required fallback before authentication or Provider execution.",
      "evidenceUrl": "https://github.com/tidbcloud/aisupport/tree/main/auto-oncall-services/docs/opts/2026-09-05-wetech-multiple-target-route-fallback.md"
    }
  ]
}
-->

- Optimization ID: `wetech-multiple-target-route-fallback-2026-09-05`
- Owner: auto-oncall-services
- Source: APID-29465 single-ticket deterministic repair
- Implementation status: local candidate, not deployed

## Bounded problem

When Wetech routing returned more than one target, strict route validation
raised `diagnostic_wetech_route_target_count_invalid`. The initial routing
boundary handled timeouts but rethrew this bounded ambiguity, causing the
analysis request to end as a service error without either response message.

## Attempted change

The initial routing boundary now sends only that exact multiple-target
validation error through the existing bounded Router fallback. It does not
start authentication or Provider execution and does not relax route validation
for any other invalid target shape.

## Current evidence

A deterministic runner regression supplies two Wetech canonical targets and
requires a successful `human_required` result with both messages, empty
requested context, no clarification simulation, and zero Provider attempts.

## Unresolved gaps

The candidate has not been deployed or verified by a post-change APID-29465
read replay, so it is not yet effective or verified.

## Next validation action

After deployment, replay APID-29465 in read mode and verify that the same
ambiguous route completes with the bounded human handoff and no Provider call.