Yes, perhaps the crash rate of drones is high, but then again there is no insight into what exactly goes wrong since this is classified. Someone did point out that they are used in very extreme environments.
I would assume that the airframe itself in drones is fairly conservative, hence relatively inexpensive which makes the whole thing expendable. Maybe that is one of the advantages the remote technology offers - you can really push it to the limits since you don't have to worry about the human pilot aboard. And since you don't engineer it to carry a human it is much cheaper. You produce many airframes and then just tweak the software - so it is possible to test many solutions in parallel.
Once the base software (i.e. algorithms) is developed for a drone the solution will be relatively easy port to a different airframe with parametrisation. Software engineers are good at that sort of thing.
The CIO at one of the companies I used to work for loved to just walk in to the datacentre and remove a few random chosen power or network cables from the panels - this is how he tested our failover really worked. I guess the same approach can be applied for the drone software. If you can complete the mission in an ideal environment - remove the pitot probe and modify the software so it understans pitch/power. Got that working? Switch off one engine and repeat. Got that working? Switch off both engines in cruise and program it to glide to a landing. Got that working? Disable the rudder actuator, or the gps receiver, or the flap motor, or the INS. etc. Yes, humans can cope with all of these issues, but the truth is the likelihood a pilot will encounter any of these in his/her flight career is probably well below zero - so we need to rely on recurrent simulator training and hope that when something bad occurs the humans can recall what they learned. A computer system, once programmed to cope with any of these situation - will never forget or become rusty. It will be able to recall the required actions with 100% accuracy.
I am also sceptical that we will ever (well at least in say the next 50 years) see a passenger aircraft which has no one 'technically' overseeing its operation on board. I can, however, imagine that we will no longer have two persons up front solely devoted to controlling the plane. There will be a 'flight manager' type role who will be able to intervene in the case of an exceptional situation, i.e. commanding the system to proceed to the nearest vs. the best airport in the case of a medical emergency, commanding the system to land immediately in the case of some cabin incident, initiating the emergency procedure in the case of some major communication failure, etc. etc. This person will not be able to directly 'manipulate the controls', merely issue overall commands. If anything is non-standard the system will seek 'flight centre' approval, if this is not possible it will revert to an emergency procedure, or perhaps seek 'human approval' from authenticated managers of other aircraft in the vicinity?
It wasn't that long ago that lifts (aka elevators) had a driver ;-)
Golf-Sierra