Gremlin - Choose one item at random

You're pretty close to your answer:

gremlin> g = TinkerGraph.open().traversal()
==>graphtraversalsource[tinkergraph[vertices:0 edges:0], standard]
gremlin> g.addV('user').property('id',1).as('1').
......1>   addV('user').property('id',2).as('2').
......2>   addV('user').property('id',3).as('3').
......3>   addV('user').property('id',4).as('4').
......4>   addV('post').property('postId','post1').as('p1').
......5>   addV('post').property('postId','post2').as('p2').
......6>   addE('follow').from('1').to('2').
......7>   addE('follow').from('1').to('3').
......8>   addE('follow').from('1').to('4').
......9>   addE('posted').from('2').to('p1').
.....10>   addE('posted').from('2').to('p2').
.....11>   addE('liked').from('1').to('p2').
.....12>   addE('liked').from('3').to('p2').
.....13>   addE('liked').from('4').to('p2').iterate()
gremlin> g.V().has('id',1).as('me').
......1>   out('follow').
......2>   aggregate('followers').
......3>   out('posted').
......4>   group().
......5>     by('postId').
......6>     by(project('likedBySelf','likedByFollowing').
......7>          by(__.in('liked').where(eq('me')).count()).
......8>          by(__.in('liked').where(within('followers')).order().by('id',shuffle).values('id').limit(1).fold()))
==>[post2:[likedBySelf:1,likedByFollowing:[3]],post1:[likedBySelf:0,likedByFollowing:[]]]

I pretty much just added limit(1) so that only the first item after shuffle is selected. It took a few executions but I was able to see both of the outputs you were looking for with this method. As I suggested on your other question, you might also use sample(1):

gremlin> g.V().has('id',1).as('me').
......1>   out('follow').
......2>   aggregate('followers').
......3>   out('posted').
......4>   group().
......5>     by('postId').
......6>     by(project('likedBySelf','likedByFollowing').
......7>          by(__.in('liked').where(eq('me')).count()).
......8>          by(__.in('liked').where(within('followers')).sample(1).values('id').fold()))
==>[post2:[likedBySelf:1,likedByFollowing:[3]],post1:[likedBySelf:0,likedByFollowing:[]]]
gremlin> g.V().has('id',1).as('me').
......1>   out('follow').
......2>   aggregate('followers').
......3>   out('posted').
......4>   group().
......5>     by('postId').
......6>     by(project('likedBySelf','likedByFollowing').
......7>          by(__.in('liked').where(eq('me')).count()).
......8>          by(__.in('liked').where(within('followers')).sample(1).values('id').fold()))
==>[post2:[likedBySelf:1,likedByFollowing:[4]],post1:[likedBySelf:0,likedByFollowing:[]]]