Regarding CPU and memory resource allocation for containers

According to my analysis of the trace, I found that there are some containers, whose CPU limits are much larger than their CPU request, in many machines. I wonder if those containers are processing offline tasks, and allowed to utilize idle CPUs in the system.
However, their CPU usages seem to indicate that they do not use more CPU than their requests, which makes me very contradictory. The trace only shows a single value for memory limits. Does this mean that the hard limit and soft limit of the container's memory are both set to the same value? Why not set different values?

Start with the referenced trace and compare the container CPU requests, CPU limits, observed usage, and recorded memory limits across machines. The issue does not name project files or tests. Done would require an explicit project-supported explanation of these resource fields and why the trace exposes memory as a single value.

Written by the indexing model from the issue text.