@fguitton if you follow the trick pointed by @stgraber you should be able to perform the procedure without container downtime.
Success 
!!!
I was able to run an lxc list and lxc cluster list on all hosts without issues.
The output of lxd sql local "SELECT * FROM raft_nodes;" is also consistent across machines.
Bonus, all my running containers are still alive and well !
Thank you very much @stgraber and @freeekanayaka for this invaluable help !
Would there be some extra sanity checks to run to make sure it’s all back on track. I saw that 4.0.3 had been released yesterday, I wouldn’t want my fate to be temped again 
Nice! You should be good now. If you hit other upgrade issues, let us know here before doing anything else, thanks.