Tom Quackenbush writes:

You lost me. If you crack the valve on an O2 bottle, even in a
vacuum, the gas will go in one direction and the bottle in the
opposite direction. Doesn't that require force?

If we assume the geometry of, oh, a conventional scuba tank, part of the
kinetic energy will be imparted to the bottle, which will acquire both spin
and translational motion, But since the "system" is simply the bottle and
contents, all it can push on, is itself. The net effect is to propel the
expelled gas and the bottle, converting the potential energy into kinetic
energy. How much of this kinesis is in the flying bottle versus the gas
depends on geometry of the exit orifice.

Consider the case where the gas output is vented out in two orifices
pointed out in opposite directions from the center of the bottle. The
thrusts are exactly opposite, so the bottle won't move. All the kinetic
energy goes into the motion of the expanding gas.

Consider the case where you can make a massless bottle dissolve instantly,
leaving the compressed gas. You just have a glob of gas that expands
forever. The work potential is all converted into kinetic energy of the
expanding gas.