Mechanics
In reducing the size, I chose to use the smaller mechanical pencil spring (12mm uncompressed, 4mm compressed) as the propulsion method. Preliminary tests with the smaller spring yielded satisfactory results in terms of projectile range and speed.
The firing rod was also size reduced, with a diametrical drop from the Rocket Launcher's 3.175mm bamboo stick to a 3mm diameter Gundam part runner. The firing stroke was limited unlike the Rocket Launcher's massive 70mm displacement to a more modest 16mm. Previous experiments in firing mechanisms showed that additional stroke length did not appreciably contribute to projectile performance.
Actuation
This design uses a paper clip bent in a box shape to act as a flat spring. There are two "arms" that connect to an obstructing rod to block the head of the firing rod, with the remaining side fixed to the barrel. Actuation is provided by a 3m MtG laminated arch that depresses the paper clip arms, lowering the obstruction and allowing the firing rod to complete the stroke forwards. A very primitive mechanism, but it offered reliability.
The initial design used a fairly long flat spring, with arms 20mm long. It worked just as well after a reduction to 13mm in length. For this design, a shorter mechanism is ideal. A shorter spring provides a longer stroke and requires more effort to depress, giving the user more of a sensation of pushing a button.
Diagrams
Perhaps the most pointless part of the post, since none of you are ever going to try to build this. Posting it here as reference regardless. Only core mechanism parts are diagrammed. The model's fairly simple that you can half-ass the rest and still have it look decent.
Prototype Testing Results
The initial prototype cycled through several rounds successfully, but there was some sign of wear where the firing rod head contacted the latch spring. There was also a significant reliability problem with the initial design, where pushing the firing rod back with any ordinance met with some interference with the latch spring. To solve this problem, another step was added to the firing rod, so the latch secured a section in the middle of the firing rod head than at the front. Stroke travel was not affected.
Ammunition
For the sticky bomb, I used a clear Gundam runner and stuck the ball of spikey death on top. From estimates, a sticky has spikes that are 1/5 of the diameter in height. I wasn't concerned with accuracy of the sticky, but more of "will it fit in this stupid launcher". A 7mm ball with 2mm spikes barely worked, as my inner barrel diameter was 11mm. Had to trim them down a little to fit better. Also, for ease of loading, I opted not to put a spike on the front of the projectile.
Prototype Demonstration (With test fire)
I really need to learn how to make videos that don't suck so bad.
Notes
In another case of "crappy TF2 modeling", it should be noted that the game model when viewed in first person vastly differs from the model viewed in third person, unequipped. The dimensions for the side plates on the ammunition drum changes. The part that I presume is an ammunition ejection port also flops sides. I suggest that one or the other be chosen for making a replica, and not both. I used the 3rd person, unequipped viewmodel for this project.
In terms of production costs, this model can be assembled for around 6 Magic: the Gathering cards and some generous amounts of cardstock. Not that you'd need one since no one else has a 12" Demoman, and this model's pretty bland, visually.