Add performance tuning section for high-throughput Runners

[ltamaster]

Summary

Adds a new Performance tuning for high-throughput Runners section to runner-config.md that documents the main JVM system properties available when users hit the server-side error:

Failed: IOFailure: Runner did not deliver reports in the configured timeout period

This error indicates the Runner is saturated (not that the operation itself hung), and until now there was no customer-facing documentation describing how to scale the Runner to handle higher concurrency or log throughput.

Properties documented

• Operation concurrency: runner.operations.maxRunning (default 50)
• Report delivery (with warning):
  • runner.reporter.sendRate (default 2s)
  • runner.reporter.sendBatchSize (default 1000)
• Micronaut HTTP client:
  • micronaut.http.client.pool.max-connections (default 50)
  • micronaut.http.client.pool.acquire-timeout (default 10s)
  • micronaut.http.client.read-timeout (default 60s)
  • micronaut.http.client.connect-timeout (default 10s)

Each property describes its purpose and when to increase/decrease it. Each subsection includes a runnable java -D... -jar pd-runner.jar example.

Design notes

• The Report delivery section opens with an admonition warning that changing these values increases the HTTP request rate/size reaching the server and can raise CPU/memory/database load on busy deployments. The defaults are explicitly recommended.
• Internal-only levers (micronaut.executors.*, micronaut.netty.*, polling rates, kill switches) are deliberately omitted — they are either risky to expose or not related to the timeout class this section targets.
• No dashboard references are included since the relevant Grafana dashboards are internal for now. Once metrics documentation is written, a follow-up can link from this section.

Test plan

• Review the rendered markdown for proper formatting (headings, admonition, code blocks)
• Verify the internal anchor #configure-java-heap-size still resolves (used in the maxRunning caveat)
• Confirm the admonition renders correctly with the docs site theme

🤖 Generated with Claude Code

[Copilot]

Pull request overview

Adds customer-facing documentation for tuning Runner performance when high concurrency/log throughput leads to report-delivery timeouts.

Changes:

• Introduces a new “Performance tuning for high-throughput Runners” section explaining the timeout symptom and what it indicates.
• Documents key JVM -D properties for operation concurrency, report batching/flush behavior (with warning), and Micronaut HTTP client pool/timeouts.
• Provides runnable java -D... -jar ... examples for each tuning area.

[Copilot]

The examples in this new section use pd-runner.jar, but earlier on the same page other examples use runner.jar / pdrunner.jar. This inconsistency can confuse readers who copy/paste commands. Consider standardizing the jar filename across this page, or add a short note in this new section clarifying that pd-runner.jar is the expected Runner jar name (or that it should be replaced with the actual jar filename the user downloaded).

[ltamaster]

@copilot apply changes based on this feedback

[Copilot]

Standardized all jar filename references across the page to pd-runner.jar in commit 2408478. This covers the proxy section (pdrunner.jar → pd-runner.jar) and the heap size/temp dir sections (runner.jar → pd-runner.jar), so all examples on the page now match.