This record chain is part of the incident-to-fix guide, which covers the metadata packet a partner collects. This page does a narrower job: name every key in the chain, show where each one is written, and mark the joins that usually fail.
The chain in one table: step, record, key, system
Read each row as one link. The key column says what connects that record to the row above it.
| Step | Record | Key that links it to the step before | System |
|---|---|---|---|
| 1. Report | Incident channel or thread | Channel name and creation time | Slack |
| 2. Triage | Incident ticket | Key posted in the thread | Jira |
| 3. Work starts | Working copy named by the engineer | Key in that name | GitHub |
| 4. Change | Saved changes | Key in each change description | Same code host |
| 5. Proposed fix | PR (pull request) | Key in the PR title | Same code host |
| 6. Review | Approvals on the PR | PR number | Same code host |
| 7. Merge | Merge into the target line of code | PR number | Same code host |
| 8. Status update | Resolved message | Message time; the key only if someone typed it | Slack |
| 9. Closure | Ticket resolution with linked development work | Key, through the links made at steps 3 to 5 | Jira |
| 10. Postmortem | Review document | Key, if the author cited it | Varies |
Two identifiers do almost all of the work. The ticket key carries the chain from step 2 to step 5, and the PR number carries it from step 5 to step 7. Everything else is a clock reading. Atlassian's documentation says that typing the work item key into the names and titles engineers give their work links that work to the Jira item in connected tools, GitHub included. In practice, the key typed by an engineer at step 3 decides whether step 9 can see the fix at all.
How each link is made, and who makes it
Not every join in the chain is equally strong. Some are made by the software, some by people remembering.
- Thread to ticket: made by routine. The on-call engineer posts the key in the incident thread. Slack's documentation says exported messages carry a user ID and a timestamp, so a key posted as a message survives as text. A key that lived only in someone's head does not.
- Ticket to working copy: made by naming. The key in a working name or change description is a convention, and the tool builds the link from it. Skip the convention and no link exists.
- PR to review and merge: made by the platform. The approvals and the merge sit on the PR itself. GitHub's protection settings can require a set number of approving reviews before a merge and can dismiss approvals made stale by new commits, but only where an administrator switched them on.
- Merge to closure: made by two different mechanisms. Jira sees the merge through the key. GitHub's own auto-close syntax is a separate feature: it links a PR to a GitHub issue only when the PR targets the default line of code, so a fix merged to a release line does not close anything through those keywords.
- Status update to everything else: made by nobody. A "resolved" message is a timestamp in a thread. It joins the chain only by time.
- Issue to postmortem: made by the author. If the review document cites the key, it joins. If it does not, the root cause and the change sit apart.
Where the chain commonly breaks
Most breaks happen at the joins people make by hand, and most of them happen late in the incident, when people are tired and the pressure is off.
| Link | Typical break | Sign it happened | Repair available |
|---|---|---|---|
| Thread to issue | Incident handled before anyone opened an issue | Thread starts well before the issue's creation time | Join by time window and channel, flagged as inferred |
| Issue to branch | Hotfix branch named for the symptom, not the key | Merged change near the incident with no key | Match by merge time and changed files, flagged as inferred |
| Pull request to review | Release branch without review rules | Merge with no approvals recorded | None; a review that never happened cannot be added |
| Merge to closure | Issue closed after the first merge | A later merge in the incident window not linked | Partial, if the later change can be tied by time |
| Status update to final merge | Resolved posted before the last fix | Final merge timestamp later than the resolved message | Yes, if both timestamps survive |
| Issue to postmortem | Review document filed by date, not by key | Postmortem cannot be found from the issue | Yes, if the document names the incident |
The fourth and fifth rows are the case that matters most for AI training and evaluation. A chain that ends at the first merge teaches a model that the first fix worked.
One incident traced through the chain
Illustrative: the times and identifiers below are fictional.
| Time | Step | Record | Key carried forward |
|---|---|---|---|
| 14:02 | Report | Private incident channel opened: card payments timing out | None yet |
| 14:09 | Triage | INC-412 created, severity 1 | Key posted in the channel |
| 14:31 | Work starts | Branch INC-412-rollback reverts a connection-pool change | Key in branch name |
| 14:48 | Review and merge | PR 1873 approved by one reviewer, merged to main | PR number |
| 15:40 | Status update | Resolved message posted | Timestamp only |
| 16:25 | Second fix | PR 1879, a retry-storm hotfix, merged to the release branch from branch hotfix-retry | None |
| 17:10 | Closure | INC-412 closed, linked to PR 1873 only | Key |
Read top to bottom, the chain looks complete until 16:25. The second fix carries no key, was merged to a non-default branch, and landed 45 minutes after the resolved message. From Jira, the incident appears solved by the rollback. A matching merge time and the changed files can tie PR 1879 back to INC-412, but that is an inferred link and should be labeled as one, never presented as part of the recorded chain. The missing-context page shows how differently the same kind of incident reads with and without the surrounding systems.
What a partner notes about the chain, and who can authorize a discussion
A partner describes the chain, not its contents. The useful notes are: which identifier is the convention at each step, whether incidents run in public or private channels, whether release and hotfix branches have review rules, and roughly what share of hotfix pull requests carry the issue key, as the platform team estimates it. Never ask for an export, a screenshot or a sample incident.
The platform engineering manager and the Jira administrator can answer those questions; neither can authorize anything. The CTO or another executive with authority over company data decides whether to discuss a license, and counsel checks customer contracts, since incident threads often quote customer details.
The company needs 50+ full-time employees at peak (contractors excluded), a documented operating history, rights to license the records and an authorized contact. Companies in the U.S., Canada and the EU qualify, as do companies in the UK, Australia, New Zealand, India, Singapore, South Africa, Latin America, the UAE and Saudi Arabia when at least 70% of the data is in English. What stays unknown until SourceX reviews it: which links survive export, whether customer data limits use, and whether there is demand for this history. To score the chain before an introduction, use the incident-to-fix completeness scorecard. Ticket-first companies can start from the ServiceNow incident page or the ITSM ticket history page; the Jira and GitHub pages cover each system on its own.
Next step
Refer a company. No cap on rewards. Default reward on a company's first qualifying transaction: the greater of $50,000 plus 5% and 15% of Actual Supplier Payout, with no cap; further qualifying transactions first agreed within 12 months of the first Closing pay 15%. Eligibility and payment conditions apply, including the USD 100,000 Minimum Transaction, and a signed Schedule A or addendum can differ. A lead, meeting or signature alone does not trigger payment.
If a client's platform team can say which key links each step and where the chain breaks, register as a partner, check the company against who qualifies, and read the program terms before introducing its CTO.