Why do concurrent writers to the same partition only get snapshot isolation instead of full isolation?

Paimon's two-phase commit protects atomicity — a snapshot appears completely or not at all — but it doesn't serialize every writer against every other writer touching the same partition. Doing so would mean forcing concurrent writers into a single queue, which would badly limit throughput for any pipeline with more than one job writing into the same partition.

Instead, Paimon allows two writers targeting the same partition to commit independently and merges the results afterward: the final table state ends up reflecting a combination of both commits, without either commit silently overwriting or losing the other's changes. This is a deliberate throughput-versus-strict-isolation tradeoff — different partitions already commit fully in parallel with no contention, so the same-partition case is really the only place this weaker guarantee shows up.

In practice this means: if you need one writer's changes to strictly happen "after" another's within the same partition, you need to coordinate that at the application/pipeline level, since Paimon itself only promises no data loss, not a specific ordering between concurrent same-partition writers.