A* can be morphed into another path-finding algorithm by simply playing with the heuristics it uses and how it evaluates each node. This can be done to simulate Dijkstra, Best First Search, Breadth First Search and Depth First Search.
A* expands on a node only if it seems promising. It's only focus is to reach the goal node as quickly as possible from the current node, not to try and reach every other node.
If you have many target nodes and you don't know which one is closest to the main one, A* is not very optimal. This is because it needs to be run several times (once per target node) in order to get to all of them.
Dijkstra is an uninformed algorithm. This means that it does not need to know the target node beforehand. For this reason it's optimal in cases where you don't have any prior knowledge of the graph when you cannot estimate the distance between each node and the target.
Since Dijkstra picks edges with the smallest cost at each step it usually covers a large area of the graph. This is especially useful when you have multiple target nodes but you don't know which one is the closest.
If we take for example 3 Nodes (A, B and C) where they form an undirected graph with edges: AB = 3, AC = 4, BC=-2, the optimal path from A to C costs 1 and the optimal path from A to B costs 2. If we apply Dijkstra's algorithm: starting from A it will first examine B because it is the closest node. and will assign a cost of 3 to it and therefore mark it closed which means that its cost will never be reevaluated. This means that Dijkstra's cannot evaluate negative edge weights.
When traversing one tree level, you need a way to know which nodes to traverse once you get to the next one. The way this is done is by storing the pointers to a level's child nodes while searching it. The pointers are stored in a FIFO way, this means that BFS needs a relatively large amount of memory in order to store the pointers. The amount of course depends on the complexity of the graph tree and the amount of nodes and/or levels.