"Consider first rock towards the top of the mantle where pressure is a less important factor than it would be at the base."
First, the pressure from overburden is important throughout the entire mantle.
Even the top of the mantle has somewhere between 10 and 70 kms of rock sitting on it and that creates huge pressure.
So, you ALWAYS have to consider gravitational compression.
"Local variations in temperature may occur, caused, perhaps by inhomogeniety in radioactive elements."
There is absolutely no evidence for this.
To be fair, almost nothing concerning the distribution of radioactivity is known for sure, so who knows? The levels of radioactivity are just educated guesses. There is no direct evidence for the levels of radioactivity currently assumed to exist in the mantle and significant contrary evidence for the levels of radioactivity that some have assumed for the core (the best educated guess for radioactivity in the core seems to be zero).
It is a fact that globally, the heat from any radioactive decay is leaking faster than it is being produced and thus there is no buildup of temperature. To account for the current rate of heat loss from the Earth, you have to add a sizeable chunk of "secular heat" (about 65% of the total heat) to that produced by the supposed radioactivity (about 35% of the total heat).
"Theory maintains that the hotter rock will be lighter than the colder rock, so, given that the material is sufficiently plastic, the hotter rock will rise, relative to the colder."
Your proviso, that the material is sufficiently plastic, is unwarranted. Mantle rock is the least plastic material on the planet (excepting the iron of the solid core) and thus is NOT sufficiently plastic.
"Theory maintains that the hotter rock will be lighter than the colder rock, so, ...., the hotter rock will rise, relative to the colder."
This is simply not true. Hotter rock will not rise, relative to the colder, even if you ignore gravitational compression.
There are these things called chemical bounds which hold all solids together. For convection to occur in a solid, quadrillions of chemical bounds must be broken and reformed. Breaking these chemical bonds is called melting the material.
Any rock where convection might occur, has to have most of these chemical bounds broken, i.e., it has to be at least partially molten.
As it is, the mantle is (everywhere) stiffer than the strongest steel. Nowhere in the mantle is it sufficiently plastic for convection to occur, even locally.
And, it should be noted that local convection can not explain the movement of the continents.
If you heat one end of a long (very wide) rock pillar to a few hundred degrees and keep the other end at zero, then convection will NEVER occur. This is because, for convection to occur those quadrillions of chemical bounds must be continually broken and reformed as one portion of the solid rock moves through another portion of it.
Unless temperatures approach the melting point of the material, the temperature difference cannot provide enough energy to break these bonds. Also, it is impossible to maintain large local temperature differences, for long periods of time, as the heat will escape by conduction. This, by itself, implies that local convection is impossible.
Although convection cannot occur, continuous massive pressure over a long period of time will cause even the stiffest steel to deform. Similarly, continuous massive pressure (provided by gravity) over long time periods, will cause some small movement of pieces of the (solid) mantle.
However, any large scale movement of one part of the mantle through another part of the mantle requires that the mantle be (at least) partially molten (as it would have been in the Mansfield's collision hypothesis). Any convection in a non-molten mantle is just a daydream.
