🛠 优化
- 导出/恢复并行化：导出与恢复任务从串行循环改为按"最大并发数"受限并发处理，大批量备份/恢复速度显著提升
- 消除重复媒体流查询：每次处理的媒体流仓储查询从 3~4 次降为 1 次
- 新入库事件早退：已具备完整媒体信息的 STRM 项直接跳过，不再排队和延迟
- 明确恢复与探测职责：恢复（纯本地）与探测（远程）拆分为独立操作，坏 JSON 降级由任务层统一编排
- 取消语义完整化：任务取消后 Emby 正确标记为"已取消"而非成功完成；进度报告单调递增
- 清理死代码与冗余：抽取计划任务基类，移除无调用方的方法与冗余字段
🐛 修复
- 修复提取任务节流失效：处理延迟移到持有并发槽位、探测之前，真正限制对远程媒体的突发压力；JSON 恢复仍不延迟
- 修复坏 JSON 降级绕过节流：JSON 损坏或恢复失败后的降级探测现在同样经过延迟和并发限制
- 修复批量恢复无限并发：提取任务中所有项目（含 JSON 恢复）统一受"最大并发数"限制，避免大量并发写入 Emby 仓储
🛠 Improvements
- Parallel export/restore: Export and restore tasks now run with bounded concurrency (MaxConcurrency) instead of a serial loop, significantly speeding up bulk backup/restore
- Eliminated redundant media stream queries: Repository queries per item reduced from 3~4 to 1
- Early exit for new items: STRM items that already have complete media info are skipped immediately without queueing or delaying
- Clear separation of restore vs. probe: Restore (local-only) and probe (remote) are now independent operations; fallback from broken JSON is orchestrated by the task layer
- Complete cancellation semantics: Cancelled tasks are now correctly marked as "Cancelled" by Emby instead of completed; progress reporting is monotonic
- Dead code cleanup: Extracted a shared scheduled-task base class; removed unused methods and redundant fields
🐛 Fixes
- Fixed ineffective throttling in extract task: The processing delay now applies while holding the concurrency slot, before probing, truly limiting burst pressure on remote media; JSON restore still runs without delay
- Fixed throttling bypass on broken JSON fallback: Fallback probing after corrupted JSON or restore failure now goes through the same delay and concurrency limits
- Fixed unbounded restore concurrency: All items in the extract task (including JSON restores) are now bounded by MaxConcurrency, avoiding heavy concurrent writes to the Emby repository
Full Changelog: v2.5.2-emby...v2.6.0-emby