When a step fails on Kubernetes build infrastructure, the build pod is cleaned up immediately. For steps that log very little, such as Harness-provided plugin steps, there is often no time to run kubectl logs, and node-level log forwarders (Splunk, Fluent Bit and similar) cannot ship the container output before the pod is deleted. This makes failures hard to debug and leaves gaps in a team's own observability pipeline.

Please add an option, at the pipeline or infrastructure level, to retain the build pod for a configurable period after a failure before cleanup. Alternatively, guarantee that step stdout and stderr are flushed to the Harness log service before teardown. This would help CI and IaCM users who run builds on their own clusters and need to troubleshoot failures that currently leave no trace.