Fix duplicate _id generation by using MongoDB-managed _id and _key-based upserts

[FestinBiju]

Summary

This PR fixes duplicate document ID collisions reported in #439 by removing custom _id generation logic that depended on timestamp + in-memory counter behavior.

Instead of overriding MongoDB _id, writes now rely on MongoDB-managed _id values and use a stable logical key (_key) for idempotent upserts.

Fixes #439

Root Cause

The previous _id strategy was not restart-safe:

• _id was derived from the current millisecond plus an in-memory counter.
• The counter reset on process restart.
• Under large chunks or high-throughput writes, collisions could occur.

What Changed

• Removed custom/manual _id assignment in ingestion and write paths.
• Updated write operations to upsert by _key instead of _id.
• Added or updated a unique index on _key (or the appropriate compound logical key where required).
• Kept writes idempotent using upsert: true.

Why This Is Safe

• MongoDB ObjectId generation avoids application-level _id collisions.
• Logical uniqueness is enforced at the database level through a unique index.
• Reprocessing identical data updates existing records instead of creating duplicates.

Verification Performed

• Replayed the same payload twice; no duplicate logical records were created.
• Bulk and concurrent write scenarios completed without duplicate _id collisions.
• Duplicate-check aggregation returned zero duplicates on _key.
• Indexes were verified as unique on _key or the intended compound key.
• Existing tests pass, along with new or updated tests for idempotent upsert behavior.

Notes on Migration / Backfill

If duplicate records already exist for _key, unique-index creation may fail until the data is cleaned up.

Recommended rollout:

1. Deploy the code path using _key-based upserts.
2. Backfill and deduplicate existing records.
3. Create or validate the unique index.
4. Monitor write errors and ingestion latency.

Risk / Impact

Low to moderate — this change touches the write/upsert path and indexing.

The main risk is pre-existing duplicate logical keys during unique-index creation. This is mitigated through the deduplication step described above.

[Copilot]

Pull request overview

This pull request removes application-managed MongoDB _id generation from the LiveTiming ingestor document model, relying instead on MongoDB’s native _id handling while keeping a stable _key to support idempotent writes/upserts and avoid restart-related ID collisions (issue #439).

Changes:

• Removed the custom timestamp/counter-based _id generation logic (sync + async).
• Updated Document.to_mongo_doc_* serialization to stop injecting _id and to document MongoDB-managed _id, while still adding _key.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[br-g]

Hello @FestinBiju,
Thanks for proposing this fix!

This issue is pretty complex though. Let's continue the discussion here: #439 (comment)