I have a few systems where this has been achieved. It’s a little messy but it works as intended.
For each queue, set up the main queue as normal and then set up a separate ‘call delivery’ queue. The first queue only has one agent - which is the second queue. The second queue (call delivery) has the actual agents you want taking calls. Agents log in and out of the second queue only.
The first queue can have your initial hold music set, and the second one can have your post-queue music. The second queue needs to have ‘queue no answer’ set to ON, and 'failover destination’s set to CONGESTION. This should mean that callers will only go through the second queue when an agent answers, and therefore they will get the correct hold music applied.
I hope that helps, let me know if you need further info.