Lakehouse Operations & Migrations — Senior ScreeningA senior-loop round: concurrency, cost, and migration judgment. Several answers require arithmetic or a tradeoff, not recall.

Two writers read snapshot N, both write data files, and both try to commit. Under optimistic concurrency, what happens to the second commit?

A table receives 20 commits a minute from a streaming writer, each adding ~10 small files. Roughly how many files does one day add before compaction?

Merge-on-read (delete/positional files applied at query time) trades what for what?

Snapshot expiration is necessary because…

A migration from a warehouse to a lakehouse plans a 'shadow + parity + cutover in waves' approach. What is the parity step for?

Which catalog choice best preserves the ability to leave a vendor later?

Write-audit-publish (WAP) prevents which failure?

A stream-ingested Iceberg table needs two-minute freshness. Which lever most directly controls the file-count bill?

Partition evolution in Iceberg lets you…

A compaction job keeps failing to commit on a table with a busy streaming writer. The most likely mechanism is…