fix(authz): bound retries for missing attribute values

[strantalis]

Proposed Changes

A single unknown FQN turned a 250-value lookup into 251 SDK calls. Split failed batches breadth-first, retain successful sub-batches, and skip missing singleton values while preserving per-resource denial behavior.

Limit failed batch probes to eight before falling back to singleton lookups. Sparse misses improve substantially; an all-missing batch is bounded at 258 calls versus the previous 251. Errors other than NotFound are returned without retries.

Layer 4 of 7 in the authorization performance stack. Review and merge from the bottom upward.

Checklist

• Added or updated unit or integration coverage appropriate to this layer
• Ran focused verification for this layer
• Documented behavior and validation limits below

Testing Instructions

• Tests verify all-known values, first/last missing values, all-missing values, and unavailable-policy errors.
• One missing value in a batch of 250 now uses at most 17 calls, verified by SDK call counters.
• Focused access v2 race tests passed.

Stack-wide validation:

• Focused race tests, the full verbose policy integration suite with the race detector, SDK README code-block tests, formatting, and service lint restricted to changes since the stack base passed with zero new issues.
• make test was attempted with Colima configured. The round-trip suite failed because its required platform server at 127.0.0.1:8080 was not running.
• make lint stopped because the configured Buf API token is invalid. Separate make go-lint reported existing repository issues; the formatting issue in the new hierarchy test was corrected.
• make govulncheck reported vulnerabilities in the existing dependencies and the installed Go 1.26.3 standard library. This stack changes no dependency versions.
• GitHub CI is still running. Keep this stack in draft pending review and CI results.

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[github-actions]

Benchmark results, click to expand

Benchmark authorization.GetDecisions Results:

Metric	Value
Approved Decision Requests	1000
Denied Decision Requests	0
Total Time	231.58811ms

Benchmark authorization.v2.GetMultiResourceDecision Results:

Metric	Value
Approved Decision Requests	1000
Denied Decision Requests	0
Total Time	120.012844ms

Benchmark Statistics

Name	№ Requests	Avg Duration	Min Duration	Max Duration

Bulk Benchmark Results

Metric	Value
Total Decrypts	100
Successful Decrypts	100
Failed Decrypts	0
Total Time	403.358065ms
Throughput	247.92 requests/second

TDF3 Benchmark Results:

Metric	Value
Total Requests	5000
Successful Requests	5000
Failed Requests	0
Concurrent Requests	50
Total Time	40.438413035s
Average Latency	403.407143ms
Throughput	123.64 requests/second

[github-actions]

⚠️ Govulncheck found vulnerabilities ⚠️

The following modules have known vulnerabilities:

• otdfctl
• service
• tests-bdd

See the workflow run for details.

[github-actions]

X-Test Results

✅ go@main-main
✅ go@main-pull-3989
✅ go@pull-3989-pull-3989
bats-test-results
✅ go@main-v0.9.0
cukes-report
✅ go@pull-3989-v0.9.0
✅ java@main-pull-3989
✅ java@main-main
✅ go@pull-3989-main
✅ js@main-main
✅ java@main-v0.9.0
✅ js@main-pull-3989
✅ js@main-v0.9.0
opentdfplatformNQABD4.dockerbuild
govulncheck-failure-8
govulncheck-failure-3
govulncheck-failure-1