Support bandwidth aggregation across multiple relay nodes

Is your feature request related to a problem? Please describe.

Yes.

A common deployment for Hysteria 2 is:

Client → Relay (middle server) → Exit server

In many cases, the relay server has limited bandwidth. For example, I may have three relay servers, each with only 10 Mbps uplink. A single live stream requires around 25–30 Mbps, so one relay server is not sufficient. Currently, I have to manually split traffic or deploy external bonding software.

It would be useful if Hysteria 2 could support bandwidth aggregation across multiple relay nodes.

Describe the solution you'd like

I'd like Hysteria 2 to provide a pluggable bandwidth aggregation mechanism.

For example:

• Configure multiple relay servers as a relay group.
• Hysteria automatically distributes packets across the available relays.
• The receiver reorders and reconstructs the original stream.
• If one relay fails, traffic is automatically redistributed to the remaining relays.
• The aggregated throughput can exceed the bandwidth limit of any individual relay.

Example:

• Relay A: 10 Mbps
• Relay B: 10 Mbps
• Relay C: 10 Mbps

Expected available throughput: approximately 30 Mbps (depending on network conditions), allowing a single 25 Mbps live stream to pass through.

A plugin architecture would also allow the community to experiment with different aggregation algorithms without changing Hysteria's core.

Describe alternatives you've considered

Current alternatives include:

• Running external bonding software (such as MLVPN or similar solutions).
• Manually splitting traffic across multiple relay servers.
• Increasing the bandwidth of a single relay server.

These approaches either increase deployment complexity or are not always available due to provider bandwidth limits.

Additional context

This feature would mainly benefit users running high-bitrate live streaming or other continuous large-bandwidth traffic through relay servers.

The key goal is bandwidth aggregation rather than failover. Existing load balancing or failover mechanisms cannot increase the available bandwidth for a single connection, while an aggregation plugin could potentially do so.