Any question answered...

what was your answer

In the top tube above the midpoint between the wheel centres.

Since we're riding on a velodrome, we have to consider the different paths taken by the wheel contact points and the CG round the bends. By maximising the height of the ballast above the line connecting the tyre contact points, we raise the CG as high as possible. The track bike broadly maintains a constant speed of the CG, and since it follows a smaller radius path than the wheel during the corners, the wheel speed is higher than the CG speed because the structure constrains all parts of the cycle to have the same angular velocity. Raising the CG increases this differential so that the measured wheel speed (which is what determines the competition outcome) is maximised. We also maximise the vertical drop of the CG on entry to the bend, and this potential energy is also converted into additional speed

If we concede the ISP required to keep @Emyr 's proposition within the bounds of the question, then for most users a position at the top of the ISP is likely to be better than anywhere in the top tube, so he was right for the wrong reason.

This being a "physics problem", we have naturally conducted the experiment in vacuo in order to eliminate aerodynamic considerations. The mass consideration remain valid even when we let the air back in, and since the centre of pressure is also way above the tyre contact line, aerodynamics only add to the increase in wheel speed mid-corner relative to the speed of the centre of pressure, since the aerodynamic drag would drop significantly due to the centre of pressure following the smaller radius path if wheel speed were held constant.

This large periodic rise and fall in wheel speed, and hence cadence, is actually a disruptive factor in pursuits and hour record attempts, which some riders handle better than others.

And now you've all made me check my working, this looks like a fun start for more advanced study, and there's even a picture for the non-readers showing a wheel speed variation of about 5% twice per lap

1 Attachment