System task are slow in conductor

[atul-X]

We are using conductor for some of our onbording workflows in fintech domain . We are facing issue where conductor system tasks are taking 3-4 for for scheduling only.
Configuration

Database persistence type.

conductor.db.type=postgres

mysql

spring.datasource.url=jdbc:postgresql://host.docker.internal:5432/conductor
spring.datasource.username=postgres
spring.datasource.password=postgres

redis queues

conductor.queue.type=redis_standalone
conductor.redis.hosts=host.docker.internal:6379:us-east-1c
conductor.redis-lock.serverAddress=redis://host.docker.internal:6379

Elastic search instance indexing is enabled.`

conductor.indexing.enabled=true
conductor.indexing.type=opensearch
conductor.elasticsearch.url=http://os:9200
conductor.elasticsearch.indexName=conductor
conductor.elasticsearch.version=0
conductor.elasticsearch.indexReplicasCount=0
conductor.elasticsearch.clusterHealthColor=yellow

Additional modules for metrics collection exposed to Prometheus (optional)

conductor.metrics-prometheus.enabled=true
management.endpoints.web.exposure.include=prometheus

Load sample kitchen-sink workflow

loadSample=true

#logging.level.com.netflix.conductor=DEBUG
#logging.lev`el.org.springframework=DEBUG

CRITICAL: Reduce workflow evaluation delay from 500ms to 100ms (5x faster)

conductor.sweep-frequency.millis=100

System Task Workers Configuration

conductor.system-task-workers.enabled=true
conductor.app.system-task-worker-thread-count=12
conductor.app.system-task-worker-poll-interval=50
conductor.app.system-task-queue-pop-timeout=100
conductor.app.system-task-max-poll-count=12

Workflow Sweeper Configuration

conductor.app.sweeper-thread-count=12
conductor.app.sweeper-workflow-poll-timeout=2000

System Task Callback Duration (keep default)

conductor.app.system-task-worker-callback-duration=30

conductor server configuration is 4 vcpu and 8Gi memory

conductor version is git:(v3.21.12)