Reduce system test sharding overhead

[cgcote]

What does this PR do?

Tunes the upstream system-test desired_execution_time using measured CI Visibility durations to minimize end-to-end workflow wall time. This is isolated from the separate Gradle clean experiment in #12785.

Selected target: 12.5 minutes (248 shards). It stays below the upstream 256-job cap while reducing the slow tail substantially.

Results

Target	Pipeline	Shards	Median	p95	Slowest	Runner time	<5m
15m control	37m43s	193	9m54s	18m25s	24m15s	1,913.7m	20
20m	39m36s	134	14m12s	25m52s	29m00s	1,849.9m	14
12.5m	29m46s	248	8m07s	16m20s	19m19s	1,977.4m	47
12.5m repeat	32m19s	248	8m07s	16m01s	19m39s	2,004.5m	42
12m	34m24s	256	7m38s	15m23s	20m58s	1,997.6m	43
13m	35m13s	235	8m27s	17m21s	22m03s	2,071.4m	31

The two 12.5-minute runs average 31m03s, 6m41s or 17.7% faster than the 15-minute control. Both also reduced the slowest shard by more than four minutes.

The 20-minute treatment used fewer runner-minutes but slowed the pipeline. The 12-minute treatment hit the 256-shard cap and regressed wall time. The 13-minute treatment reduced startup count but regressed wall time, p95, maximum, and runner cost. Together these bound the fastest observed target at 12.5 minutes.

The planner consumes historical duration data from system-tests time-stats.json and uses first-fit-decreasing allocation; it does not currently query live CI Visibility data. CI Visibility is used here to measure and select the target.

Inspired by DataDog/dd-trace-rb#6439, while keeping setup-cost changes outside this PR.

Validation

• All six measured experiment workflows completed successfully.
• The selected 12.5-minute treatment was repeated successfully.
• Workflow YAML parses successfully.
• git diff --check passes.

[dd-octo-sts]

🟢 Java Benchmark SLOs — All performance SLOs passed

Suite	Status
Startup	🟢 pass

SLO thresholds are defined here based on automatically generated metrics. A warning is raised when results are within 5% of the threshold.

PR vs. master results

Scenario	Candidate	master	Δ (95% CI of mean)
startup:insecure-bank:iast:Agent	13.88 s	13.85 s	[-0.4%; +0.9%] (no difference)
startup:insecure-bank:tracing:Agent	12.81 s	12.92 s	[-1.6%; -0.2%] (maybe better)
startup:petclinic:appsec:Agent	17.10 s	16.99 s	[-0.2%; +1.6%] (no difference)
startup:petclinic:iast:Agent	16.88 s	16.91 s	[-1.1%; +0.7%] (no difference)
startup:petclinic:profiling:Agent	16.54 s	16.71 s	[-1.8%; -0.2%] (maybe better)
startup:petclinic:sca:Agent	17.05 s	16.97 s	[-0.5%; +1.4%] (no difference)
startup:petclinic:tracing:Agent	16.14 s	16.08 s	[-0.6%; +1.3%] (no difference)

Commit: 1f529938 · CI Pipeline · Benchmarking Platform UI

---

Load and DaCapo benchmarks can be triggered manually in the GitLab pipeline. Results will appear in the Benchmarking Platform UI after completion.