Just seen that question, so I am years late to the game. It still makes sense to put an answer here because, well, the pandemic is not over to the degree where it will not significantly affect attendance, and even after that there are reasons for remote practice.
We've practised throughout the hard lockdown in our country (about half a year) using Jamulus, free software with a long history (at one point of time, it used to be called llcon instead). It uses a centralised server architecture focused around UDP packaging of Opus-encoded data (these days, WebRTC might make more sense), where the server is responsible for creating the individual mixes (everyone has their own mixer for managing the sources) and sending them out again.
Since it is audio-only, integrating a conductor requires some imagination. The central server should be in "network vicinity" to the clients (which can mean that it may be better situated near a network interconnection point 100km away than in the same town as the participants) and have sufficient power for handling all the clients. If you don't want to depend on a public server, renting an on-demand server of sufficient power and connectivity will make sense because they tend to have better connectivity than a private user.
Running the server is what requires technical savvy (Linux offers itself), setting up the clients (Windows, Linux, MacOSX, by now also iOS and Android) is somewhat straightforward, but low-latency setups (avoiding WiFi, preferring soundcards that can reach low latency, typically ruling out bidirectional USB microphones that would otherwise be perfect) do warrant some experience. Note that a traveling expert can provide help here in one-on-one scenarios that are less problematic than group settings regarding pandemic safety measures.
For fast pieces and mixed user latencies it may be helpful if someone will inject a fixed rhythm from an arranger keyboard or a similar "deaf drummer" contraption: that helps avoid "creeping slowdown". Headphones are mandatory since echo compensation is not done for speed/efficiency reasons.
The individual tracks can be recorded on the server and a mix generated and distributed from it afterwards: voice separation is much better than can be achieved in a live setting which may make this a helpful tool outside of pandemic reasons.
For getting your feet wet, it can be easiest to start installing a client, then connecting to some public jam session on a public server. Be sure to have read up on stuff well enough not to disrupt the session if you can avoid it. Developers and users tend to be friendly, however.
A different software would be "Sonobus" which has a peer-to-peer architecture and consequently has increasing network and performance requirements for every client as the number of users grows: for Jamulus only the server needs to scale up to the number of users regarding network and processing power.
Some people claim to get along better with JamKazam particularly regarding the complexity of the setup, however it requires working with a central server that has a fee scheme for membership.
All of those solutions are audio-only. For some public Jamulus sessions it is not uncommon to have a Jitsi session for video only in parallel, but in my experience it affects the audio network performance too much to be really desirable.