Memory leak persists in large dataset runs (~100MB increase per run) - Follow up to MCOL-6064Memory leak persists in large dataset runs (~100MB increase per run) - Follow up to MCOL-6064

Issue Overview
The customer reports that the memory leak reported in MCOL-6064 appears to be unresolved when running against larger datasets. While previous fixes reduced the leak rate significantly for smaller operations, the customer is observing a memory jump of approximately 100MB per run in their current testing environment.

Environment

• Architecture: 3 Node Cluster
• OS: RHEL 8
• MariaDB Version: 10.6.15-10
• ColumnStore Version: 23.02.16
• Data Size: 911 GB

Customer Observations
The customer conducted a sequence of 14 runs using the attached scripts.

• Run 1 Memory Usage: ~400 MB
• Run 14 Memory Usage: ~1740 MB
• Observed Behavior: Memory usage increases by roughly 100MB between runs.

Attachments

• Latest Logs (See attached)
• Reproduction Scripts (See attached)

Notes for Engineering/Product

• PM Context: It was previously noted that the leak was reduced by 100x+ (down to 1.5MB over 10 mins).
• Discrepancy: The customer's data indicates a regression or a scaling issue where the leak expands significantly (100MB/run) with larger datasets (911 GB) and repeated execution.