The reason is logged as the cause:
No ManagedConnections available within configured blocking timeout ( 50000 [ms] )
You have set a maximum of 100 connections and a blocking timeout of 5000ms.
If you increase the nummer of parallel throughput the getConnection might wait for the next free one.
As the handling is not fair it might happen that one is waiting longer as others and hit the timeout.
Also if your transactions run for longer and block the connections.
You might check this in the jmx.console, there are several counters for the connection pool.
In fact you have to increase the max-pool-size.