SQL execution does not change, but consistent read get higher

Question

Hi,
Our SQL query execution does not change, but consistent reads get higher in our test enviroment but query looks fine in our development and production environments.
In the Development instance the trace is:
STAT #18446744071526492680 id=1 cnt=1 pid=0 pos=1 obj=7151 op='TABLE ACCESS BY INDEX ROWID MEMBER_CMS_SITE_ACCESS (cr=4 pr=0 pw=0 time=153 us cost=3 size=68 card=1)'
STAT #18446744071526492680 id=2 cnt=1 pid=1 pos=1 obj=7152 op='INDEX UNIQUE SCAN MEMBER_SITE__MEMBERID_SITEID (cr=3 pr=0 pw=0 time=104 us cost=2 size=0 card=1)'
So - we read 3 blocks from the index and then 1 from the table ( to make 4)
In the Test instance the trace is:
STAT #18446744071411593144 id=1 cnt=1 pid=0 pos=1 obj=7151 op='TABLE ACCESS BY INDEX ROWID MEMBER_CMS_SITE_ACCESS (cr=112 pr=0 pw=0 time=2820 us cost=3 size=70 card=1)'
STAT #18446744071411593144 id=2 cnt=1 pid=1 pos=1 obj=7152 op='INDEX UNIQUE SCAN MEMBER_SITE__MEMBERID_SITEID (cr=3 pr=0 pw=0 time=90 us cost=2 size=0 card=1)'
We read 3 blocks from the index but the table needs 109 more which cannot possibly be right.
It looks like we are applying UNDO and those 109 are the work needed to do so.
I have tried flushing the shared pool, and killing some sessions that could possibly be the cause... but with nothing in V$TRANSACTION
I could understand this if there was a long-running transaction that's created a lot of dirty blocks, but there are no transactions at all... so it's not that.
Thanks

Answer 1

The SQL is very simple with only one predication in the where cluse and that is indexed,just like
"select * from table where a=:b0"
Column a is indexed. But ORACLE choose FTS not INDEX SCAN because of out-of-date stats maybe.
So I updated the stats expecting to see ORACLE will choose INDEX SCAN. The fact is that ORACLE will not change existing FTS to INDEX until I flush the entire share pool.That's the problem.