What you were expecting:
When a request gets canceled it should allways dispatch a FETCH_CANCEL action, to decrease the loading count.

What happened instead:
When two requests that have {cancelPrevious: true} get fired simultaneously, the FETCH_START for both gets dispatched, one of them gets canceled, but the corresponding FETCH_CANCEL never gets dispatched. This makes the loading count become locked at a greater than 0 value forever.

Steps to reproduce:
dispatch a crudGetOne and a crudGetList simultaneously (can be for diferent resources).

Environment

• Admin-on-rest version: 1.4.*