One hit per second for tracking purposes would be reasonable.
Google Analytics is a good case study for it. Google Analytics has "events" that you can grammatically send via JavaScript and would allow Google Analytics to track user actions such as start and stop. You could also send "continuing to play" events every minute of audio. From their collection limits and quotas:
analytics.js:
Each analytics.js tracker object starts with 20 hits that are replenished at a rate of 2 hit per second. Applies to all hits except for ecommerce (item or transaction).
So the Google Analytics API would allow tracking to the half second. If Google Analytics allows it, it must be manageable even at huge scale.
Another way to look at is is in terms of overhead. If you are already streaming music to the users you have to have a powerful enough server for that. A ping track a minute is going to use an infinitesimal amount of bandwidth compared to streaming music. It may require an additional connection to your server. That would require configuring your server to accept more connections, but connections are generally not a super scarce resource.