I dislike pwm due to the current ripple it pulls from the line, and that ripple will be pulled out of your xantrex's input capacitors btw..
I suppose you could have the 600 volt line run in series through the water heating elements, use relays to short the elements out by default.
the panels will still be run at their MP point.
That's a possibility ! Normally shorted except when the batteries go to Absorb or Float or EQ (Voltage regulation stage).
Then, when you un-short part of that series resistor, the battery voltage will drop below set-point voltage and the controller will be forced into dropping its input voltage and go back to MPPT and of course heat up the dump load.
Not nearly as voltage controlled as the parallel dump load but this might have some merit. Or, might be just fine ?
One thing, Caradaman about some of these controllers is that they are bidirectional in nature. They create the very input voltage from the battery through the controller that they are making the PV source see at the controller's input terminals. So, under certain conditions, the controller can have reverse current from the battery, through the controller and back out its input terminals into your dump load. It's hard for it to drive the source though unless, for some reason the controller makes its input voltage go above the Voc of the PV array or higher than the rectifier rating of the wind or hydro circuit. But a resistor dump load, it can certainly drive that way. This would hopefully make the controller either disconnect or reduce its input voltage.
Also, a PWMing dump load will, as Joestue said, cause ripple on the controller's input capacitors and heat them up unnecessarily. You can fix that by adding a diode with its cathode connected to the controllers' PV positive input terminal. That takes care of the negative current from the battery and also help that ripple current in the caps.
It doesn't help the fact that the PV array is not putting out maximum power point voltage and power into that dump load though. That is the job of the controller and its output (MPP tracked) goes into the battery side (the low voltage output). The controller doesn't make the maximum power flow through to the dump load that is now on the input.
The system could do something about that though if there was a current sensor on the dump load and software to control the PV voltage so that actually happened but that is quite a different system.
It's kind of an interesting problem though I must say but there may be something I am not thinking of here.
boB