Oracle RAC Load Balancer Not Working?

I'd like to talk about some interesting behaivor with Oracle RAC load balancer.We have Oracle RAC 11.1.0.6 - 2 nodes.

We noticed sometimes the number of sessions { select count(1) from v$session; } was not divided equally between nodes. Also load average was very different between node #1 and node #2.Now we define service explicitly.

srvctl add service -d MYORCL -s MYSERVICE -r MYORCL1,MYORCL2
srvctl config service -d MYORCL -a
MYSERVICE PREF: MYORCL1 MYORCL2 AVAIL: TAF: NONE

After we change connection string of corresponding application(s) the spread will be more equal:

select service_name,inst_id, count(*) the_qty
from gv$session
where service_name = 'MYSERVICE'
group by service_name, inst_id

SERVICE_NAME INST_ID THE_QTY
---------------------------------------------------------------- ---------- ----------
MYSERVICE 1 68
MYSERVICE 2 79

Although it's not round robin algorithm, the distribution looks more equal. It seems explicitly defined service of Oracle RAC has better load balancing approach comparing to the default service (created with db by default).

But there is some downside with more 'equally' distributed load balancing. Consider a client program that is running the following SQL once a 5 minutes:

select ...
from A
inner join B on ...
inner join C on ...
inner join D on ...

where the tables B, C and D data is changing all the time.Once load balancing is distributed more 'equally', the query above will arrive to node #1 and to node #2 more equally. It will cause more interconnect traffic, because Oracle RAC will need to synchronize requested blocks all the time, with much high rate.