Thank you 256bits, for correcting my apathy ever so gently. I really do appreciate your challenge.
The bouyant force will allow the gas with greater density to sink in the room, thereby causing the gas with lower density to rise. This will, to some extent, set up convection currents. Interestingly, the convection currents themselves may generate areas of low pressure in the room if there are obstructions in the flow (Bernouli's Principle), although I do not know whether they would be significant.
Low density gas may have two distinct causes. First, the fire will warm the gases closest to it and will cause thermal expansion in the gas. This is the cause of ordinary convection currents which we observe in campfires, fireplaces, and wood stoves even, which currents are the cause of vacuums at the base of the fire.
Second, different gases have different densities because of their chemical makeup. The density of O2 at NTP is 1.429 kg/m^3; density of CO2 is 1.977 kg/m^3 (source:
http://www.engineeringtoolbox.com/gas-density-d_158.html). This means that the CO2 byproduct will tend to stay low in the room while O2 is forced toward the ceiling (which may explain why fire fighters are trained to vent rooms by breaching high).
So CO2 is denser than O2 and there would be a partial vacuum created by combustion of carbon based materials. But, I conclude that the expansion due to heat is greater than the potential loss in density due to the change in chemistry. Any vacuum caused in a container or room which is almost entirely sealed has as its major cause the subsequent contraction of cooling air.
The fire, in a backdraft situation will cause the volume of gas in the room to expand, forcing air and smoke out of any opening (cracks, holes, gaps, etc.). This process will continue until the fire runs out of one of the three necessary items (heat, oxygen, fuel). (Note, this may seem counterintuitive in an ordinary fire where a flow exists because we see air being forced in continually at the base of the fire.) While structure fires do not generally lack oxygen, it could happen in a sealed room where heat is contained and plenty of fuel is available. Air would be pushed out until the fire reached its critical point at which time the density of air in the room would stabilize. From that point, the room would begin to cool down which would produce lower density, creating a partial vacuum. The vacuum would be filled with air from the "hallway," which presumably is denser, colder, and has more oxygen.
The colder, denser, oxygen-rich air would expand as it heats up inside the room. This may be partially responsible for the puffing, but would not likely sustain the condition. The introduction of oxygen into the room would allow further combustion, producing additional heat, returning the fire to its previous state (hot, but lacking oxygen).
I've written too much already. I apologize for my many words as I think out loud.
Thoughts?