Explain the lifecycle of a write operation in Qdrant (WAL, segments, optimizers)?

A write in Qdrant moves through several stages before it's fully reflected in a queryable, optimized index — first ensuring durability, then applying the change to in-memory/segment structures, then eventually being folded into the collection's ongoing background optimization process.

Every write is first appended to the Write-Ahead Log, which guarantees durability — if the process crashes right after acknowledging a write, the WAL can be replayed on restart to recover any changes that hadn't yet been fully applied to segment storage.

The change is then applied to an active segment, making the point immediately available for queries even before any background optimization runs; over time, as many small writes accumulate into many small segments, a background optimizer process merges them into larger, more efficient segments and rebuilds the HNSW index accordingly, which is what keeps query performance from degrading as a collection experiences ongoing write traffic.