From what I understand, a long-running checkpoint could block the creation of a checkpoint cursor. I believe it should open the last successful checkpoint but after looking at the code, it seems that it will return EBUSY if there is a checkpoint running and this EBUSY is handled and retried internally.

See here and here where we set EBUSY and the internal loop here

What needs to be done:

• Create a test where we open a checkpoint cursor while a checkpoint is running
• Add stats / logs when a checkpoint cursor cannot be opened