Sandbox list pagination omits newer destroyed sandboxes that remain accessible by exact ID

Question

Hey guys. I've been trying out Railway sandboxes and have run into the following problem. Any help would be greatly appreciated!

## Summary

The Railway Sandbox collection query and dashboard omit recently created

sandboxes, even though those sandboxes still exist and are accessible through

an exact-ID query.

This appears to be a listing/pagination or indexing issue rather than a sandbox

creation failure.

## Environment

- Observed: 2026-07-30, approximately 09:03–09:23 UTC
- Railway TypeScript SDK: `railway@3.6.0`
- Railway CLI: `5.8.0`
- Authentication: project token
- Network isolation: `ISOLATED`
- Idle timeout: 15 minutes
- All affected sandboxes were subsequently destroyed normally

I can provide the project and environment IDs privately.

## Reproduction

Two independent processes created sandboxes in the same Railway environment.

Each process created:

1. One sandbox sequentially
2. Two additional sandboxes concurrently

Every create request had a unique environment nonce, so the mutations were not

byte-identical.

One affected sandbox was:

`ad1b5077-06fd-42d3-a404-fcbfeaf64265`

An exact lookup for that ID succeeds and returns:

```json
{
  "id": "ad1b5077-06fd-42d3-a404-fcbfeaf64265",
  "status": "DESTROYED",
  "createdAt": "2026-07-30T09:04:00.391Z"
}
```

However, the same sandbox is absent from the environment's `sandboxes`

connection.

We tested both:

```graphql
sandboxes(environmentId: $environmentId, first: 100)
```

and:

```graphql
sandboxes(environmentId: $environmentId, last: 100)
```

Both returned the same 100 records and the same page information:

```json
{
  "hasNextPage": false,
  "hasPreviousPage": false,
  "startCursor": "059ba302-18c6-4fec-af2d-2f5a763fed57",
  "endCursor": "5bd4943e-068c-4625-905d-2c3eb02e4e6b"
}
```

The newest returned record was created at:

`2026-07-30T09:03:53.205Z`

That is earlier than the omitted sandbox, which was created at:

`2026-07-30T09:04:00.391Z`

The omitted sandbox also does not appear in the Railway dashboard.

## Expected result

One of the following should happen:

- The newer sandbox appears in the collection; or
- Pagination reports another page through `hasNextPage` or

`hasPreviousPage`.

`first` and `last` should not return the same capped page when newer accessible

records exist outside it.

## Actual result

The collection returns exactly 100 destroyed sandboxes, omits newer accessible

records, and reports that no additional page exists.

## Impact

Sandbox creation and execution work, but the dashboard and collection query

cannot be used for lifecycle history or auditability. It also prevents us

from reliably reconciling run-owned sandboxes against Railway's inventory.

Please let me know if you need any more info on this