My $.02 worth involves riding twice with S-Health cycling, the first of those two times with CycleDroid operating in parallel (on the same Galaxy S5 phone). In the first ride (27.5 miles in 108 minutes), S-Health ended up with good screen data at the end (i.e., distance, average speed, etc.) *but* the audio announcement of average speed was about 1/3 too low (e.g., when I knew I was making a 15 mph average from my cycle computer, it would say something like 10.4 mph). I don't know how it could screw up the obviously correct internal data in the translation to the audio output, but it did. CycleDroid simply dropped out for several miles and then clicked in again (no pauses in this entire ride). Neither of these results was acceptable.
For the second ride I used S-Health only and the distance was exactly 20 miles (I've measured it many times with independent GPS as well as my calibrated cycle computer) and around mile 8 or 9, S-Health just started outputting the "distance to go" as one mile more than it really was (so the distance already covered was exactly 1 mile less). Also, the time of ride changed by -4 minutes 21 seconds. So when I finished my 20 mile ride, it said 19 miles had been covered and the average speed given was 9.4 mph but my cycle computer gave a (correct) speed average of 15.9 mph. Also, the total distance shown on the screen time (along with the time of ride) was incorrect... so it wasn't just a problem with the audio output.
Both apps are non-reliable, in my opinion, based on these experiences, and I won't use either of them again. Very hard to understand how this (S-Health) software could become so buggy, if it was good before (this was my first use of either of these two apps).