Hi Railway Support,

I'm investigating an intermittent connectivity issue between Shopify and my Railway-deployed application, and I'd like to get help from someone familiar with Railway's networking infrastructure.

Shopify reports that requests to my application sometimes fail because the connection/handshake times out.

Initially, my setup was:

Shopify → Cloudflare → Railway

I suspected Cloudflare might be involved, so I disabled the Cloudflare proxy completely (DNS-only) for several days. The issue still occurred, so the affected requests are now going directly:

Shopify → Railway

During one of the failures, Railway's network logs showed an ingress TCP flow with the following characteristics:

flowState: partial
l4Protocol: tcp
direction: ingress
peerKind: internet
l4LatencyMs: 5
packetCount: 11
byteCount: 5750
dropCause: UNKNOWN_104

The source of the connection was a Shopify IP, and the flow occurred at approximately:

2026-09-04 20:32 UTC

The application is deployed in Railway's US East region.

The particularly interesting part is that the flow was marked as partial and had:

dropCause: UNKNOWN_104

Could someone please clarify:

1. What does UNKNOWN_104 mean?
2. What does flowState: partial indicate?
3. Does this indicate that Railway dropped/terminated the connection, or could it mean that the remote side stopped sending/closed the connection?
4. Was the connection successfully established at Railway's edge?
5. Is there anything in Railway's networking/edge infrastructure that could cause an intermittent TCP/TLS handshake timeout like this?

I've already removed Cloudflare from the request path and the issue continues, so I'm trying to determine whether the failure is occurring at the Shopify → Railway network layer or somewhere within Railway's infrastructure.

I'd appreciate any insight from someone who can interpret these network-flow fields or investigate the corresponding flow internally.

Thank you.