Troubleshooting
Reproduce reporting mistakes, inspect incorrect results and check corrected queries against boundary cases.
Literal strings and tuple representations
Check patterns, prefixes, numeric text order and lossy tuple encoding.
Boolean filters and conditional counts
Separate literal codes, truth conversion, counted branches and zero fallback.
Exact and signed numeric reports
Test precision, truncation, magnitude and optional known extrema.
JSON report types and levels
Preserve field presence, array kind, direct depth and object values.
Query stages and local ranking
Order filtering, aggregation, history and quotas at their correct scope.
Sets, recursion and comparison context
Check exact sets, distinct graph nodes, collations, affinity and unknown ordering.
Byte decoding and preserved text
Diagnose replacement, boundary and normalisation losses before accepting text records.
Interpreting workbook cells
Resolve extracted cell types, references, cache evidence and merged coordinates.
Payload admission and retained meaning
Hold ambiguous members and preserve declared value kinds and extensions.
Constructing the destination schema
Check destination column layout, write constraints, generators and export shape.
Loading related records
Diagnose actual foreign-key failures, regenerated links and replacement effects.
Write failures and indirect effects
Inspect transaction scope, conflict handling and triggered effects after a load.
Binary bundles and relocated files
Check decoded byte budgets, destination paths and references without losing source identity.
Migration reruns and publication
Verify step identity, rollback material and committed publication state.
Diagnose dates changed during spreadsheet import
Check the original representation and the target field’s interpretation before importing the full file.
Which stock counts disagree with the movement ledger?
Reconcile opening stock and signed movements with a physical count before correcting an inventory balance.
Which invoices disappear when joined to customers?
Find issued invoices whose customer key has no matching master record.
Why do invoice totals grow after joining payments?
Aggregate multiple payments per invoice before calculating invoice and payment totals.
Which active customer accounts share a normalised email?
Create a review list for duplicate email keys without automatically merging identities.
Which confirmed room bookings overlap in time?
Detect conflicting pairs using half-open time intervals and exclude cancelled bookings.
Which closed exceptions have since reopened?
Find currently open cases with a prior closure instead of counting every historical reopen event.
Which import rows are missing required values?
Test missing email and missing amount separately, keeping zero amounts distinct from NULL.
Which imported dates are ambiguous or impossible?
Reject missing, noncanonical and impossible date-only inputs before importing them.
Why does a department disappear from the open-work board?
Keep every configured department visible, including teams with only closed tasks or no tasks.
Why does a skills join inflate team headcount?
Count people rather than person-skill pairs when preparing a staffing overview.
How do matching invoice numbers connect the wrong supplier?
Reconcile a supplier invoice with its supplier-and-number identity instead of the number alone.
Why does NOT IN hide equipment awaiting inspection?
Find equipment with no matched inspection despite an unrelated inspection row with a missing equipment ID.
Why are tickets with an unknown state missing from review?
Preserve unknown ticket states in a not-confirmed-closed review queue.
Why is an unused budget shown as a missing balance?
Treat the absence of matched expenses as zero spending, while keeping a missing expense amount distinct.
Why does a valid discount percentage produce zero?
Avoid integer division before calculating a discount in integer cents.
Why do draft bookings make cancellation rates look lower?
Use the agreed completed-or-cancelled cohort and return unknown when that cohort is empty.
Why do repeated check-in scans inflate attendance?
Count each member once per session while preserving repeated scans as separate events.
Why does SUM(DISTINCT) lose legitimate stock receipts?
Distinct quantities are not distinct receipt identities. Sum accepted unique receipt records.
Why is the average receipt price wrong for purchased units?
Weight each unit price by the number of units purchased, rather than giving every receipt equal weight.
Why does the latest price date show an older high price?
Select one complete price-history row instead of combining independent maxima into a nonexistent record.
Why are two years of October sales merged?
Group by year and month together when comparing reporting periods.
Why do UTC settlements appear on the wrong Hong Kong day?
Convert the declared UTC timestamp to the fixture’s UTC+8 business clock before grouping its date.
Why is a noon pickup counted in two shifts?
Use start-inclusive, end-exclusive reporting windows to avoid double counting shared boundaries.
Why does a zero-order month disappear from the demand trend?
Use an explicit reporting calendar so a complete zero month differs from a missing period.
Why do two cash events share the same running balance?
Order individual cash events explicitly and use a row-based running frame.
Why is one joint sales leader missing from the ranking?
Choose a tie-preserving rank when the business rule recognises all joint leaders.
Why does today’s dispatch report use next month’s address?
Limit address history to the report date before selecting the newest eligible row.
Why does today’s rate change an earlier purchase amount?
Match each purchase to its approved synthetic rate effective on the purchase date.
Why does UNION drop a receipt from another branch?
Retain branch identity when stacking independent receipt exports.
Why does a debit increase the cash reconciliation total?
Apply the export’s credit/debit direction before calculating net account movement.
Why does spare stock in another warehouse hide a local shortage?
Compare each warehouse with its own threshold before making a local replenishment list.
Why can a positive final stock balance hide an impossible issue?
Inspect the running stock path rather than only the end-of-period balance.
Why is stock released on its first unusable day?
Match the comparison operator to an explicit first-unusable-date policy.
Why does an overnight booking have a negative duration?
Subtract complete timestamps rather than only their hour and minute components.
Why do individually valid parties exceed shared-room capacity together?
Calculate simultaneous seat demand from start/end events instead of testing each party alone.
Why is a two-room session marked feasible when only one room is free?
Require every configured room dependency to match the supplied availability snapshot.
Why can matching totals conceal incorrect imported records?
Compare values by stable identity and detect extra target rows as well as missing source rows.
Why does numeric conversion merge distinct customer IDs?
Keep text identities intact when leading zeros have business meaning.
Diagnose a webhook signature mismatch
Check the exact received bytes and the correct endpoint secret before changing verification rules.
Read Retry-After before repeating a rate-limited request
Interpret a seconds delay or HTTP date and coordinate requests that share the same quota.
Resolve a local time repeated by daylight saving
Keep a named zone and the chosen offset when a local clock time can mean two different instants.
Recover an attachment transfer with an expired link
Reauthorise a download for the same file identity without turning a private link into a permanent public copy.
Validate cached API data with a resource version
Match a validator to the stored representation before reusing a 304 response.
No matching resource.