Strikes me as rather clever, actually. I'd expect the rate of absences to converge down on some stable value from the (high) base rate, so you'd probably need to do some experimentation to figure out what the stable rate is. Subsidizing a low monthly payment with penalty fees might end up being a better business model in the long run, depending on what the numbers turn out to be. But paying penalty fees feels like a failure where continuing to pay a monthly fee doesn't, so I'd expect it to be effective at encouraging people to work out.
The downside is that I'd expect a number of people to be shamed into dropping their memberships altogether. Shouldn't hurt the business model as long as it's not everyone, but it does change the expected income math.
Related to the recurring topic of akrasia and anticipated near-mode losses, here's an article about "Gym-Pact," an arrangement whereby people precommit to pay penalty fees if they don't stick to their planned workout schedules.
In other words, they aren't charging customers money in exchange for a service, nor for violating an agreement associated with a service... rather, they are charging money as a service.
Had I encountered this in fiction, I would have considered it satire.
(I'm being somewhat glib here, admittedly: in this "experimental" phase, they are giving people free gym memberships as part of the deal, and using the penalty fees to pay for the memberships. But that doesn't sound like the ultimate business model.)