Using hyperledger fabric discovery service with fabric-gateway

What is the best way to go about specifying several peers that client application could contact for them to start a transaction on a channel when using fabric-gateway?

Examples I've found in docs rely on connection to a single peer, which I find problematic, since if it goes offline the application won't work. I thought about distributing requests manually application-side, but I am not sure it is the intended way to go about it.

// Create a Gateway connection for a specific client identity
gw, err := client.Connect(
    id,
    client.WithSign(sign),
    client.WithClientConnection(clientConnection),
    // Default timeouts for different gRPC calls
    client.WithEvaluateTimeout(5*time.Second),
    client.WithEndorseTimeout(15*time.Second),
    client.WithSubmitTimeout(5*time.Second),
    client.WithCommitStatusTimeout(1*time.Minute),
)

It might help to think of your client application connecting to the Gateway service as similar to a Web browser connecting to a specific Website. There is a single logical connection to a single logical endpoint. That does not mean that your client application (or Web browser) must stop working if a single Gateway peer (or Web server) goes offline.

This section from workshop material in the Fabric samples repository discusses some of the approaches that can be taken to ensure high availability:

https://github.com/hyperledger/fabric-samples/blob/main/full-stack-asset-transfer-guide/docs/ApplicationDev/01-FabricGateway.md#production-deployment-of-fabric-gateway

To summarise:

• The gRPC connection itself is a logical construct and the network connection(s) it encapsulates can fail-over from the client side between different endpoints, either using default behaviour based on single DNS entry mapping to multiple IP addresses, or using a custom policy.
• The gRPC connection will transparently attempt to reestablish failed network connections, and network connections can be directed to different physical peers using server-side load balancing to achieve high availability without the client having to worry about it.

The Gateway service within the peer uses the discovery service internally to help select the best peers for transaction invocations, and to manage fail-over during the transaction flow. Your client application should only be using Gateway peers from a trusted organisation (ideally exposed on a single endpoint address) so should have no need to interact with the discovery service to discover other network peers.