First, autonomous drones are those that require no external intervention by a pilot or operator beyond the initial programming of the flight; they work entirely on their own once you hit the go button. Remotely Piloted Air System is a better description (even though I don't like it!) of what goes on with the larger platforms such as Watchkeeper.
On the airworthiness front, 'suitability for safe flight' depends on the level of 'safe' you are trying to achieve. There is an ongoing debate across the industry about equivalence (ie an equal level of AW assurance with manned flight), as the current risks from drone ops are self-evidently 3rd-party only. That position will change when RPAS (or possibly autonomous drones?) start carrying passengers, as is now being explored in UAE, because the systems will need to meet the relevant weight-related CS standards applicable to CAT platforms. Until then, the use of intelligent routing as a means of mitigating the 3rd-party risks appears to be both ALARP ('practicable' not 'possible'...) and proportionate.
What really does need to be better handled is the professional aviation knowledge of those operating Watchkeeper and the like. The report on the Bastion accident pointed clearly to this. Whilst the root cause of the engine failure was the blocked oil pipe, an individual witth better knowledge and training might have understood the implications of rising temperatures more quickly and protected the engine thereafter. Easy to say with hindsight etc, and I am not trying to point fingers here. Similarly, the Boscombe fog accident may have been avoided had people (at all levels of the procurement and operating processes) thought about the potential pitfalls of flying a system designed for Israel in the prevailing weather conditions of N Europe. All it needed was a line in the RTS to say 'landing in visibilities less than X metres is not permitted'.