The tracks are simply too fine to synchronise all the individual heads. Even if head A is driven to cylinder 1,234,567 there's no guarantee that head B will be at the same position.
The cylinders are very narrow on modern drives are spaced only a few nanometers apart - typically around 75 nm on a state-of-the-art drive. No matter how careful the design, the quality of the materials and the quality of the parts, you can't align a part to that precision and keep the alignment despite vibration, temperature changes, etc.
You can take the active head and have it follow the cylinder (even allowing for the cylinder wobbling because the platter isn't perfectly central on the spindle - there's no way you can mount a platter on the motor and keep the center aligned to 50 nm), by using an active tracking controller, which constantly senses the position of the head in relation to the track, and making microadjustments. Indeed, that's how modern hard drives work.
So, is it possible to have each head on it's own control arm. Yes, it is. It's been tried, but it is utterly impractical. The individual arms each need their own actuator and control processor. The arms create vibration which interfere with each other's motion, etc. There were a few commercial products that worked this way; they were fast, but ludicrously expensive, power hungry and unreliable.
Shortly afterwards, RAID (redundant array of inexpensive disks) was invented. Now you could combine individual disks to get effectively a higher performing, bigger disk. What's more these individual regular disks were much cheaper, than a "super-disk".
The multi-actuator drives died, never to be heard from again. RAID was subsequently rebranded "redundant array of independent disks", as the price was no relevant (or correct).