Add configurable Kubernetes client rate limiter (QPS/Burst)

[fengcone]

Why do you need it?

The controller uses client-go default rate limiter (QPS=5, Burst=10), which is too small for production. This causes API throttling and poor performance.
No configuration option exists without code modification.

How could it be?

Allow configuring QPS and Burst via command-line flags or Helm values.

Default deployment values should be higher than client-go defaults

Other related information

None

[devguoo]

+1 on making this configurable.

For k8s-heavy controller workloads, the client-go defaults (QPS=5, Burst=10) are usually too conservative once you have multiple concurrent sandbox lifecycle events. Even if APF/server-side throttling protects the API server, the low client-side cap becomes the first bottleneck.

A couple implementation details that would make this much easier to operate in production:

• expose both values as controller flags and Helm values
• log the effective QPS/Burst on startup so operators can verify what is actually applied
• document a tuning rule of thumb (for example: increase gradually and watch client-side throttling / 429 patterns)

An explicit knob here is much better than forcing operators to patch code or rebuild images just to tune for cluster size.