Heres a better closeup of the PCB, what is the component attached to the ballast case with leads? Transistor maybe?
The thing with longer wires is the input rectifier (4 diodes in a bridge rectifier configuration, all encapsulated in one common package)
And based on the only few components there and the swirling you describe, I have the impression, than it is only a simple HF inverter (similar to fluorescent ballasts, only a bit scaled up), so no extra intelligence...
If that is the case, I won't use there the MH's, as the swirling arc, touching the arc tube wall could locally overheat the quartz there.
With HPS it does not pose as much problems, as the arc is running way cooler, so the PCA could handle it's heat pretty well, even when it touches it.
The MH's operate at way higher arc temperature, so I won't be so sure even with the ceramic arc tubes...
The intelligeht ballasts I was writing about have at least four transistors (the hypothetical minimum with power regulator combined with polarity chopper, no PFC; never seen such type), but most likely you would see 5 transistors (one is in an active PFC, four in the bridge configutration for combined power regulation and polarity chopping), the most complex one have six of them (1x in PFC, 1x in power regulation, 4x in polarity chopper).
All of them contain the input rectification and filtering, mostly with an active power factor correction (but that is not needed for the main ballasting function, but it make it easier in upstream power distribution and because it can quite precisely stabilize the DC output, it make the downstream power regulation simpler in the control part)
Following is the buck regulator DCDC stage, responsible for converting the DC intermediate voltage from the rectifier to the voltage required by the arc, while controlling the power delivered to the arc. The control circuit monitor the arc voltage and current and adjust the delivered power. It makes a basic, or more advanced diagnostic on the parameters, mainly in order to recognize lamp faults and eventually identify the lamp type and so adjust the power parameters accordingly (in multiwatt MH ballasts).
And then follow the bridge circuit, what swap the lamp polarity each few ms, in order to prevent electrolysis effects.
The last two functions are sometimes integrated into one H-bridge stage, what regulate the power alternatively in both branches. The advantage is (obviously) simpler and cheaper ballast, the disadvantage is worse trade-off between resulting arc current ripple long polarity swap dead time, so the current crest factor is not really unity. And because of the required trade-off, it does not allow to serve multiple wattages with a common HW.
When separate circuits are used for each part, quite a good filter could be placed between the power conversion and the polarity chopper, so the polarity swap gap could be really small (1us and below), while the filter could be made slow, so provide very low ripple. Such arrangement is capable to serve power ranges more than 1:4 with a common hardware, so it is only matter of control SW configuration to reconfigure the setup for different lamp wattages, eventually the SW could identify the lamp wattage, so the user only insert the wanted lamp type and the ballast drive it correctly. These are usually offered as "35-70-150W" units for store displays, where the display designer could swap the lamp wattages according the actual arrangement need, without worrying about configuring the ballast box...
