Your discharge rate is too high.
You are drawing 10A from a 5000mAh battery. Most NiMh batteries are rated for discharge at a fraction of the mAh rating. The ones I used to work with were rated for discharge at .2 times the mAh rating. So, your battery pack shouldn't be discharged at more than 1A by the rules I used to work with.
When discharged at too high a rate, NiMh batteries go bad. The internal resistance increases everytime you discharge the battery too fast. That increase in internal resistance is permanent. Eventually, the internal resistance will be so high that you can't make effective use of the battery.
The battery still has its full capacity. If you slowly discharge a NiMh battery whose internal resistance has gone up due to mishandling, you will measure near its rated capacity. If you try to pull a higher current, though, the voltage will drop. This is the effect you are seeing.
The only fix that I know of is replacing the damaged cells
Back in the mid 1990s, Motorola brought out the Visor model portable two way radio. It was a fist sized radio capable of transmitting at 5Watts - the first radio that small to be that powerful.
The Visor was delivered with a 600mAh NiMh battery.
The company I worked for sold hundreds of the things. Everybody loved them - until the batteries started dieing.
Motorola replaced the 600mAh batteries with some "improved" batteries that delivered about 700mAh in the same physical size.
These died even faster than the originals.
We investigated, and found that typical battery analysers would give the "dead" batteries a clean bill of health. According to them, the batteries were fine.
The radios, however, would shut down and restart if the bad batteries were used.
I tested them with a rig I built out of a DAQ, LabView, and a current shunt, and found that the internal resistance of the batteries was ridiculously high. This made it impossible to draw high current from the bad batteries.
More investigation turned up the recommendation from the NiMh cell manufacturer to never discharge them at more than 0.2 of the rated capacity.
The Visor drew about 2A when transmitting at 5W. That's like 20 times the manufacturer's suggested maximum.
The solution was to use the optional high capacity battery for 5Watt operation, or just replace the small batteries when they died - most of our customers needed high power and small size, so they just budgeted for a LOT of replacements.
The company I worked for back then had good contact with Cadex. At our request, they added an internal resistance test for NiMh batteries to the C7000 so that we had a quick pass/fail for checking the batteries. That feature is still in the C7000 to this day.
I actually tested several batteries to destruction by simulating typical usage patterns. I ran the simulation at 5W with the 600mAh batteries, and they died. The same test at 1W didn't kill the 600mAh batteries. The larger batteries all survived the 5W tests.