Concurrent shared sentence-transformer inference races (tensor size mismatches)

[ksaurabhAparavi]

Concurrent encode() calls on the shared NomicBert SentenceTransformer model race and intermittently trigger tensor size mismatches during instance processing. Serialize wrapper encode() and raw shared-model access. Fix prepared; PR to follow.

[github-actions]

🤖 Internal: Discord sync marker

Auto-managed by the Discord notification workflow. Stores the linked Discord message ID. Do not edit or delete.