What caused this
When Luminous receives an on-hand update from a connected WMS or 3PL (such as a TrackStar stock pull), it consolidates the new total across any internal bin-location records for that SKU. Historically, this was a two-step write: Luminous would first zero out the old bin-location rows, then write the new count into the target location. Each step was committed to the database separately.
During the brief gap between those two commits, any process that read the SKU's on-hand — including the fulfillment-order auto-generation system — would see a total of zero. The auto-generator treated this as a genuine shortage and either skipped the order or fired a low-stock alert, even though full stock was about to be restored in the very next write.
What was fixed
The two writes are now wrapped in a single atomic database operation: the new quantity is written to the target location first, and the old bin rows are zeroed after — all within the same transaction. A concurrent read can no longer observe a zero total: at worst it sees a brief over-count (the new value plus the not-yet-cleared old value), which does not trigger a shortage.
A daily retry sweep has also been added. It scans for orders that have no fulfillment order despite having fulfillable stock, and re-queues them for auto-generation. The sweep is idempotent — it excludes orders that already have a fulfillment order and guards against duplicates at every stage — so it is safe to run continuously.
Who is affected
This issue affected accounts that use an external WMS or 3PL stock-pull integration (for example, TrackStar) where Luminous receives absolute on-hand updates rather than incremental adjustments.
Accounts that manage inventory purely within Luminous (no external stock pull) were not affected.
What to do if orders are still stuck
Once your account has this fix enabled, the daily retry sweep will automatically pick up any previously stuck orders. You do not need to take manual action for recent orders.
If you believe older orders were affected and have not been retried, contact Luminous support — the sweep can be run for a wider historical window on request.