Crazy idea...
Rather than the traditional FFXI style sommoning,, what if the Summoning magic had incremental effects?
For instance:
SMN begins to summon Shiva. Total casting time = 30 sec
- 0:00: Casting begins and a large spherical field (see: Limsa's opening story) appears. snow begins to fall inside field and ice elemental attacks are enhanced.
- 0:05: A friendly Ice Elemental appears and begins to attack the mob.
- 0:10: Enemies within the field suffer the Frost effect and Paralyze.
- 0:15: A second friendly Ice Elemental appears and begins to attack the mob.
- 0:20: Party members within the field gain Ice Spikes and enBlizzard
- 0:25: A third friendly Ice Elemental appears and begins to attack the mob.
- 0:30: All Party and Mobs within the field are transported to Shiva's realm where Shiva appears and unleashes Diamond Dust. When the sequence is complete, the field and all it's effects end and the game returns to normal.
*Note: The final sequence would be a client-side experience and only be visible to party members within the field, all others would see the 'bubble' frost over and become opaque until the final damage part of the animation on the mob. (that is to say, outsiders won't see either Shiva's realm or Shiva herself)
I would make this magic rather fragile so that if the summoner gets hit or moves, the field and its effects immediately end and the spell must be restarted. This would also force tanks to work very hard to keep mobs within the field and yet away from the summoner. Also the summoner's emnity could increase gradually as the spell is being cast, thus making it more difficult to maintain the spell.
The notion of a summoned monster perpetually out and following a person around is much better suited for a Beastmaster (pet) job. In contrast, Avatars are extremely powerful demi-god beings and it makes more sense for the player to visit their realm than vice-versa.