Mutex locks under high load are blocking history syncer/trappers

Steps to reproduce:

1. install and use monitoring system
2. try to scale up using proxies
3. use built-in functionality like aggregated items
4. history syncers become at 100% load

Result:
Up to 1/3 of the Mutex locks will fail of the valuecache and configuration cache. For example for 5 minutes:

Strace output will be full of the:

11414 17:14:50.190574 futex(0x7fcde87bf140, FUTEX_WAIT, 2, NULL) = -1 EAGAIN (Resource temporarily unavailable) <0.000007>

Mostly high count of trapper/history syncer processes are useless on high loaded instance due to the locking problems. Dedicated hardware is not loaded at all - so vertical scale is not happening.

Expected:
The amount of processes allowed to be launched should correlate with the ability to handle workload on high performance hardware.