When did Optimizer use index FFS in queries having Max(datetime)?

Assume a Table MYTABLEA1, having date type column for example datetime:

select max(datetime) from MYCHEMA.MYTABLEA1;

Row count: 1384

 

Plan
SELECT STATEMENT CHOOSE
	2 SORT AGGREGATE
		1 TABLE ACCESS FULL MYCHEMA.MYTABLEA1

 

select max(datetime) from MYCHEMA.MYTABLEA2;

Row count: 8139221

Plan
SELECT STATEMENT CHOOSE Cost: 442,815 Bytes: 8 Cardinality: 1
	3 SORT AGGREGATE Bytes: 8 Cardinality: 1
		2 PARTITION RANGE ALL Partition #: 2 Partitions accessed #1 - #26
			1 INDEX FULL SCAN (MIN/MAX) NON-UNIQUE MYCHEMA.MYTABLEA2 Cost: 442,815 Bytes: 2.491.053.600 Cardinality: 311.381.700 Partition #: 2 Partitions accessed #1 - #26

 

I quoted folowing from Julian.

The indexed columns must have a NOT NULL constraint

The table does not need to be analysed

http://www.juliandyke.com/Optimisation/Operations/IndexFullScanMinMax.html