Thanos Receive behind load balancer

[LukaszWasko]

I already have one Thanos Receiver which works fine with the rest of components.
To increase availability I'm trying to properly configure my two Thanos Receive instances behind load balancer (haproxy, round robin) with or without hashring.
Prometheus Agent is configured to send via remote write to terminated IP service address on load balancer (so far one Receiver is active, the second is backup):

remote_write:
  - url: "http://thanos-receive.contoso.com:19291/api/v1/receive"

My questions are:

1. Do I have to use hashring on receivers behind LB?
2. Or should I remove load balancer and just add two remote_write (two receivers) on two prometheus agents (scraps the same targets)?
3. Should I use roundrobin on LB or only one active receiver node?
   I have spent a lot of time searching this on Internet and I am really confused.

Thanos Receivers version:

thanos, version 0.24.0 (branch: HEAD, revision: 7fa248981cd98d2cd1e86b06854957d5b9a213fd)
  build user:       root@e4d6a4c7a9e7
  build date:       20211222-15:45:23
  go version:       go1.16.12
  platform:         linux/amd64

thanos-receive1.contoso.com conf (so far without hashring):

/usr/bin/thanos receive --objstore.config-file /etc/thanos/s3.yaml \
  --tsdb.path=/var/lib/thanos/receive \
  --grpc-address=0.0.0.0:10907 \
  --http-address=0.0.0.0:10904 \
  --remote-write.address=0.0.0.0:19291 \
  --label='receive_replica="0"' \
  --log.level=debug \
  --log.format=json

thanos-receive2.contoso.com conf (so far without hashring):

/usr/bin/thanos receive --objstore.config-file /etc/thanos/s3.yaml \
  --tsdb.path=/var/lib/thanos/receive \
  --grpc-address=0.0.0.0:10907 \
  --http-address=0.0.0.0:10904 \
  --remote-write.address=0.0.0.0:19291 \
  --label='receive_replica="1"' \
  --log.level=debug \
  --log.format=json

Proof of HA Concept to achieve - now I have only 2 x grafana behind LB and 2 x Prometheus Agents. The rest of the components are just one node each for now: