Today the socket-timeout is used the same for all operations. This works great for single key operations and other "simpler" requests. However, operations like getAll, putAll, prepare tx among others can take much longer and will likely hit the timeout and be retried with no benefit.

Also note that ISPN-16657 talks of reworking timeout to not perform retries, which would also change this behavior a bit, by at least not repeating the same getAll/putAll over again which can cause more overloading of servers. Even with this change we will still want a way to increase this timeout for these operations.

We could automatically increase the timeout by a scaled value of key or values, but this is probably too opaque for users. We may need to have an override of some sort on a per method invocation or maybe something like Flags that can be set on the RemoteCache itself.