What would happen if we switch to a merge‑on‑read strategy? What are the pros and cons of switching?

Switching to a merge‑on‑read (MoR) strategy changes how writes and reads are handled. Writes are appended as delta files without rewriting existing data, so write latency and storage overhead drop dramatically. Reads, however, must merge base files with delta files on the fly, which can increase query latency, especially for large scans.

Pros:

• Lower write amplification and faster ingestion.

• Better support for incremental updates, deletes, and upserts.

• Enables time‑travel and point‑in‑time queries.

• Reduces storage costs for append‑heavy workloads.

Cons:

• Higher read cost due to on‑the‑fly merging.

• Requires more metadata management and compaction jobs to keep read performance acceptable.

• Complexity in tuning merge thresholds and compaction frequency.

• Some engines may need additional configuration or support for MoR.

Overall, MoR is ideal for workloads with frequent small updates and large append volumes, but you must balance read performance with the benefits of reduced write cost.