query optimizer questions

Why do the following three queries use different scan types under 7.1.2?
Under
6.5.3 they all used index (and my system ran MUCH faster, since I often
use the trailing %)

I've done a vacuum analyze.
I installed 7.1.2 on RedHat 6.2 using the RPMs.

anderson=# explain select value from s_n_grantor where value='foobar';
NOTICE: QUERY PLAN:

Index Scan using s_n_grantor_key on s_n_grantor (cost=0.00..823.58
rows=247 width=12)

EXPLAIN
anderson=# explain select value from s_n_grantor where value like
'foobar';
NOTICE: QUERY PLAN:

Seq Scan on s_n_grantor (cost=0.00..8178.46 rows=247 width=12)

anderson=# explain select value from s_n_grantor where value like
'foobar%';
NOTICE: QUERY PLAN:

Seq Scan on s_n_grantor (cost=0.00..8178.46 rows=1 width=12)

EXPLAIN
anderson=# \d s_n_grantor_key
Index "s_n_grantor_key"
Attribute | Type
-----------+-----------------------
value | character varying(80)
btree

anderson=# \d s_n_grantor
Table "s_n_grantor"
Attribute | Type | Modifier
-----------+-----------------------+----------
doc | character varying(16) | not null
entry | smallint | not null
value | character varying(80) |
Indices: s_n_grantor_key,
s_n_grantor_pkey