recover

Load persisted CRDT state from the DurableStore for all exporters.

Call once at startup, before any calls to spans.export, metrics.incrementSum, logs.export or logs.merge, or WarpOtlpBridge.drain — and never concurrently with any of them.

"Never concurrently" is not politeness. WarpLogRecordExporter.recover deliberately stays outside that exporter's write mutex, because an un-recovered exporter's segment numbering starts at its construction defaults: a logs.merge racing this allocates a segment number the persisted index already uses and overwrites a live key, in either serialization order. A lock there would order the writes while suggesting a safety it cannot deliver, so this contract is the safety.

Idempotent: a second call simply re-reads and re-decodes the same bytes.