Troubleshooting guide

Why X Likes Still Appear After Cleanup (And What To Do Next)

Seeing old likes after a cleanup pass does not always mean the cleanup failed. In practice, operators run into three realities at the same time: device cache lag, delayed propagation, and a small set of true residual IDs that still need another pass.

If you treat all leftovers as failures, you overreact and rerun blindly. If you ignore all leftovers as “cache,” you risk leaving real signal conflicts in place. The right move is a verification-first troubleshooting sequence.

1. Separate visibility lag from residual likes

Use this decision model first:

• Likely cache lag: old like appears on one device/browser only, but not on another.
• Likely propagation lag: item disappears and reappears briefly across views during the same hour.
• Likely residual: same like ID remains visible across multiple checks after follow-up pass.

This removes guesswork and keeps your follow-up pass focused.

2. Recheck with a sampled verification pass

Do not run a full rerun immediately. Start with a sample of historical like IDs from your archive and classify each sample result:

1. already gone,
2. removed during verification,
3. inconclusive (auth/rate/network).

If “already gone” dominates, you are mostly seeing stale UI/counter behavior. If “removed during verification” dominates, schedule one controlled follow-up pass.

3. Validate auth freshness before blaming the engine

Many “stuck likes” cases are stale auth headers, not logic failure. Refresh browser session headers before the follow-up run and validate them once before action calls.

A cleanup pass on stale auth creates false negatives that look exactly like residuals.

4. Use a bounded retry strategy for residuals

For verified residual IDs, run a bounded pass with delay and retry backoff. Do not loop indefinitely.

• small targeted batch only,
• wait on rate-limit responses,
• capture unresolved IDs for manual review.

This protects account stability and keeps evidence clean.

5. Cross-check with multi-view validation

After follow-up pass, validate in this order:

1. desktop logged-in view,
2. mobile app view,
3. alternate browser session.

A “clean in 2 of 3 views” pattern usually indicates short propagation lag. Recheck after 12-24 hours before escalating.

6. Keep a runbook note for support and audit

Store one short troubleshooting summary per run:

• sample size and classification counts,
• follow-up batch size and results,
• unresolved IDs (if any),
• recommended next check window.

That makes support responses faster and prevents duplicate troubleshooting cycles.

Quick response template

“Cleanup pass completed. A subset of likes may still appear due to cache/propagation lag. Verification sample is running now. We will recheck in 12-24h and only escalate unresolved residual IDs.”