Global memory consolidation remains pending and cannot be reclaimed

Open 💬 0 comments Opened Aug 5, 2026 by qianhaoq

Problem

Memory extraction completes, but global Phase 2 consolidation remains pending indefinitely and cannot be reclaimed.

Observed behavior

  • Stage 1 output records exist, including records selected for Phase 2.
  • The global consolidation input watermark is ahead of its last successful watermark.
  • The global consolidation job remains pending.
  • The pending job has no active lease, no retry time, and no reported error.
  • Fresh non-ephemeral root sessions with memory enabled do not reclaim the job.
  • Starting or restarting the background app server does not change the job state.
  • raw_memories.md, MEMORY.md, and memory_summary.md are not refreshed.
  • General runtime and database integrity checks report healthy.

Expected behavior

A stale pending Phase 2 job should be reclaimed automatically, or transition to a failed state with an actionable error and supported retry path.

A supported command or UI action to retry or safely rebuild global memory consolidation would also help.

Related

Similar to #23129, but this report focuses on a pending global job with no active lease or retry that fresh root sessions cannot reclaim.

View original on GitHub ↗