Module recovery 

Recovery procedures for corrupt shards, partial writes, and missing DB.

This module implements the runtime recovery procedures that restore a bones project to a consistent state after:

• Partial/torn writes (process crash mid-append)
• Corrupt shard data (bit flips, truncation, invalid content)
• Missing or corrupt SQLite projection database
• Missing or corrupt binary cache files
• Locked database (retry with timeout)

Recovery Philosophy

• Deterministic: same input → same recovery action, every time.
• No silent data loss: corrupt data is quarantined, never deleted outright.
• Fast common path: torn-write repair is the typical case (truncate last incomplete line). Complex cases (quarantine, rebuild) are rarer.
• User-facing messages: every action emits a diagnostic so operators know exactly what happened and why.

Structs

HealthCheckResult

Result of a full project health check.

RecoveryReport

Report from recovering a corrupt or partially-written shard file.

Enums

RecoveryAction

The action taken during recovery.

RecoveryError

Errors that can occur during recovery operations.

Functions

auto_recover

Run a full health check and auto-recovery on a bones project directory.

open_db_with_retry

Attempt to open a SQLite database with retry and timeout for lock contention.

recover_corrupt_cache

Recover from a corrupt or missing binary cache by deleting it.

recover_corrupt_shard

Recover a corrupt shard file by scanning for the last valid event line and quarantining corrupt data to a backup file.

recover_missing_db

Recover from a missing or corrupt SQLite projection by triggering a full rebuild from the event log.

recover_partial_write

Recover from a partial write (e.g., crash mid-append).