Inventory_sync batch stuck in "processing"

GraphQL API
1

We’re hitting the same inventory_sync “stuck in processing”.

We’re a 3PL building an automated pipeline that needs to push absolute (replace-style) inventory quantities from an external vendor system into ShipHero on a ~30-minute cadence, for one of our customer accounts.

What we submitted (a minimal 1-row test file, quantity 0, to a SKU that already had on_hand: 0, so it’s a genuine no-op):

CSV hosted at a URL we control:
#sku,quantity,action,reason,location
M1WHISM,0,replace,ACC sync test - no-op verification

mutation {
  inventory_sync(data: {
    customer_account_id: "QWNjb3VudDo5OTg4MA=="
    url: "https://middleware.larimarlogistics.com/acc-shiphero-sync-test.csv"
    warehouse_id: "V2FyZWhvdXNlOjExNjAxOQ=="
  }) {
    request_id
    complexity
    sync_id
  }
}

Response:

{
  "data": {
    "inventory_sync": {
      "request_id": "6aba8b059156f37ba0d77f4d",
      "complexity": 400,
      "sync_id": "6aba8b059156f37ba0d77f4e"
    }
  },
  "errors": []
}

Polling inventory_sync_status since — no movement at all:

{
  "data": {
    "inventory_sync_status": {
      "data": {
        "id": "6aba8b059156f37ba0d77f4e",
        "status": "processing",
        "error": null,
        "success_count": 0,
        "error_count": 0,
        "finished_count": 0,
        "total_count": 1,
        "created_at": "2026-09-28T15:43:01+00:00",
        "updated_at": "2026-09-28T15:43:02+00:00"
      }
    }
  }
}

Before trying inventory_sync, we also tried inventory_add and inventory_replace directly on this same SKU/warehouse/account, and both were rejected with “One of the following fields is required: location_id or lpn_id” — this account appears to be on Static Slotting with no bin locations assigned to these SKUs, so those two mutations aren’t usable for us without first building out a whole bin-location-management flow we don’t otherwise need. inventory_sync (with its “assign to unassigned bin if location is omitted” behavior) was the one path that looked workable for our use case, so this stuck-batch issue is currently a hard blocker for us.