I don't think anyone has yet put a live HORNET style source routed mixnet out there. Roasbeef talks about it in issues on the LND github related to BOLT 12 messaging.

This type of routing scheme is extremely opaque to observation because you can't tamper with it without terminating its transit. It's what LN uses under the hood for the links in a chain of payments, called Onion TLV (type length value).

The challenge of source routed mixnets like Indra and what HORNET would be like is sybil attacks and congestion mitigation. The base commitment to set up is a cost that can't be avoided, and by design clients will be preferring established and known relays, so it is not possible to pretend to offer service, gather a million sats and then abscond. You have to spend nearly that much just to get started anyhow. Good service is rewarded with loyalty.

The per-time payments are tiny and start tiny when testing out the reliability of relays. Clients also propagate node addressess so they leave out the ones they don't like, and this also helps feed back into the service quality.

Congestion mitigation is a separate problem. It can and will happen that for a sustained period a lot of packets get sent into one little knot of nodes and the buffers will get full and for a while latency and throughput appear to have fallen, when in fact everyone has just jumped on the same chair when the music stopped. So in the protocol there is constant signalling wherever possible to let clients know the traffic levels and prefer underutilised ones.

We are close to having our project ready to test on real live networks and not just proving the protocols work and messages pass around correctly.

Bitcoiners will be increasingly vulnerable as the value continues to rise. There is now businesses actively engaging in building correlations between locations and addresses.