[danpalmer]

On the contrary, I just wouldn't experiment with it.

Experiments get left on, subscriptions get left lying around, people leave, documentation doesn't get written for early prototypes, and so they end up costing a lot of money. $25/mo isn't a lot, but finding out 4 years later that it's still going after that week you were interested in Haskell is an expensive mistake.

In small companies the processes for getting sign off for these sorts of expenditures are often not developed enough, and so doing it is a pain for an experiment. In big companies it's easier in some ways but might come with more box ticking necessary to get it to happen, or more business justification.