Okay, I have tried the dynamometer but results are not good because of lack of armature. So I decided to use
femm to simulate the situation, slightly altering a coilgun example with LUA script (this is the reason the files are still called "coilgun").
Of course, jcw0752 experience is right, and the acceleration greatly increases with gap reduction. I will attach the graphics because I believe they are interesting. The femm file and lua script are attached at the bottom.
The simulation has axial symmetry, so only half plane is modeled:
The armature is all-important, including closing the gap at the other side of the plunger. Otherwise the plunger doesn't accelerate as per the formula. Here is where my simplistic model broke down: I used the formula for the inductance of whole air coil, while things are much more complex than that: the plunger is closing the magnetic circuit, compressing all the energy inside the gap. By the way, I also assumed that the current was turned off when the plunger started moving: that is clearly wrong. The current stays on, and replenishes the energy inside the gap, in an ever decreasing volume, so the plunger gets more and more energy as it advances.
All in all, the equations in the posts above work only when the plunger starts to move, and the gap is very large. They are valid not much longer.
So here is the femm computed acceleration versus gap size (force in newtons, gap in inches), showing that hard earned experience beats idle theoretics every time:
The coil characteristics, turns, current, etc. are in the femm file and are more or less irrelevant: the 1/g^2 curve is clear enough.
I will try and modify the reasoning above, using a better flux estimate, to arrive at the correct force relation. The speed question too, with this new situation.
This is a very instructive and subtle problem
Edit:Okay, the problem was in the flux computation part. One cannot use the flux of an air core coil. A complete magnetic circuit has to be considered.
Assuming the armature has zero reluctance (that's the reason it is there), the only paths of reluctance are the plunger and the air gap. The reluctance of the air gap is \$\frac{g}{\mu_0 A}\$, where g is the length of the gap. And the reluctance of the plunger is \$\frac{\mathit{l}-g}{\mu A}\$. The magnetomotive force is N·I, and the reluctances are in series, so the flux is:
\$\Phi \ = \ \frac{NI}{\frac{g}{\mu_0 A} + \frac{\mathit{l}-g}{\mu A}} \ = \ \frac{\mu_0 N I A}{g + \frac{\mathit{l}-r}{\mu_r}} \ \approx \ \frac{\mu_0 N I A}{g}\$
From there, the approximate flux is \$B \ = \ \frac{\mu_0 N I }{g}\$ and, introducing that into the energy formula, we arrive at the force: \$F \ = \ \frac{\mu_0 N^2 I^2 A}{2g^2}\$, with the gap length instead of the coil length in the divisor.
So the trouble was with taking the flux of a whole air coil, instead of a mixed air/plunger coil. The correct force is inversely proportional to the gap length, as jcw0752 said.