Loothing

This release is the final hardening pass from the live 2.0.30-2.0.33 session-flow fixes. It focuses on reconnect behavior, duplicate loot detection, and keeping council updates efficient under raid-size bursts.

Fixed

• Reconnect restores are safer and less confusing. Master looters no longer reload into a stale "remote" council or observer roster, council members keep the candidate data they are allowed to see while waiting for the live ML sync, and non-privileged observers do not restore stale private response/vote data from cache.
• ML Observer Mode behaves correctly after reload. If the master looter is configured as an observer, the council table can still open from cached session state during reconnect, but the ML remains blocked from casting council votes until live state verification completes.
• Hidden responses now hide rolls everywhere. Roll values now follow the same visibility rules as responses in council rows, sorting, and results views.
• Roll-won loot stays tied to the correct boss. Encounterless roll-win detections no longer get pulled into the next unrelated boss session, and same-item repeats across back-to-back encounters no longer collapse into the previous encounter's buffered row.
• Duplicate and stacked tradeable items are counted more accurately. Bag snapshots and post-kill scans now count stack quantities, not just occupied bag slots, and same-item bag scans compare against already represented session/buffer copies before creating a new row.
• The Loot Picker handles cold item-cache updates correctly. Rows whose filter result was based on incomplete item info re-evaluate when Blizzard's item cache finishes loading, while any checkbox the ML touched stays preserved.
• Picked loot keeps its trade-window details. Items selected in the Loot Picker now carry item ID, source, and remaining trade time into the started session.

Performance

• Candidate updates coalesce before broadcast. Multiple pending updates for the same item and candidate now collapse to the newest snapshot before sending, reducing redundant candidate traffic during response, gear, and roll bursts.
• Council update batches have bounded latency without elevated priority. Response-update batching no longer slides indefinitely under steady traffic, and flushed candidate updates use normal priority so they do not crowd lifecycle messages.
• Reconnect sync coalescing remains bounded. Multiple reconnect requests still share one gathered session snapshot and one group sync broadcast instead of fanning out full-state whispers.