Claude's memory feature has broken long‑form continuity and corrupted multi‑document workflow = unusable = months of flowing conversation in product development lost

Has anyone else noticed that the new memory system in Claude Code has severely reduced its ability to maintain multi‑document continuity? We went from conversation flow, with history recorded to allow Claude to easily move from one topic to a new window and connect to a topic. I am not a super Google Claude user; this is several hundred in $tokens and over 200 hours in development working on Claude chat and Claude code.