This totally depends on the voltage available across that distance, and the current. Arcs (a plasma channel) has an intrinsically very low resistance, so can be generalized as a short while it exists.
Low voltages (less than around 50vDC) have little ability to sustain (and "draw") an actual arc, because the ignition (ionization) distance must be so small. If high currents are involved at low voltage, spot-welding can happen however.
If one of the batteries were to become disconnected for 10ms from it's contact while outputting 25W, because of a knock, would an arc form?
If that battery is "outputting 25W" then it is in circuit, so it is possible. But this depends on how the batteries are arranged and the load impedance, inductance, capacitance, etc. So the immediate voltage differential with respect to the disconnected terminals is the arcing voltage.
- If that voltage were 0v due to another battery being in parallel, then nothing would happen.
- If that voltage were 4.2v @25W (thus 5.95A), it might not visibly spark at all due to the very low voltage; however repeat this many times and eventually the connection will be degraded (metal oxides formed) or possibly spot-welded.
- If that voltage were 10,000v @25W (thus 2.5mA) due to a large inductive load, it would definitely spark (and be able to draw an arc to at least an inch long.) 2.5mA is a small amount of current however, so the arc wouldn't be very hot and is unlikely to melt the contacts directly unless capable of arcing for an extended period. The arc would still oxidize and corrode the terminals however.