FX insert VS FX send resource usage

It's hard to tell,  tenfoot, if you're fully satisfied by the answers or not. It sounds like your question is subtly related to two different performance-related concerns, one being the overhead of routing and mixing, and the other being the effect processing.
 
In the case of a single effect on a bus, the 6 audio streams are mixed into one and processed by a single pass through one instance of the effect. In analogue terms, the signals are mixed onto an aux bus and passed into a single input on an effects box and the output is mixed back in on an aux input or channel. Then the output of that bus is mixed with the rest of the project and sent out the main bus.
 
In the case of effects on 6 tracks, separate audio streams are passed separately into 6 copies of the effect, each requiring a pass of an audio stream through the algorithm. The CPU used for the effect is thus six times in the previous case. Then the six effect outputs are mixed into the rest of the project. In analog terms, you would have to buy 5 more boxes to do that processing using channel inserts. The analogue vs DAW reasoning is somewhat similar: you want different parameter values in each of the 6 effects.
 
The routing and mixing overhead is virtually the same in the two cases. You could argue that there is an extra bus required in case 1, like the aux bus on a mixer, but the actual CPU used by the existence of that bus is negligible. At most it requires an extra step in summing the audio stream from that bus into the main mix, whereas in case 2, you could sum all audio streams at once into the main bus mix. But in technical terms, mixing audio streams is negligible compared with processing audio streams through an effect.
 
To complete your experiment, you would need to run your project with no effects, then add the one effect to see if that effect is using enough CPU to be measurable. Then your 6-effect case would be easier to interpret. In other words, first find an effect that uses significant CPU, then try six of them. You could also separate your measurement of using the extra bus or not, by running the project with and without the extra bus, but with no effects, to see if that difference is measurable.