That card is a mad gamer person confection
Insane, just insane.
It appears that the big heatpipe and cooler sticking out the back of the card is for the memory, not the core. If it's for the VRMs then it would actually be useful for mining, whereas if it's just on the memory chips themselves then it's not any use for mining.
I clock my memory at 275 MHz. Normally the minimum is 300 MHz but under load the cards run at 1,000 MHz (base - if the memory is DDR4 then the total bandwidth will be 4,000 MHz, but forget that for a moment because mining is NOT heavy on memory bandwidth).
This cooler has let the original owner run his memory clock at 1,250 MHz - for games, this will be a worthwhile mod. Won't help with mining though, and the memory runs very cool regardless if you clock down to around 300. The standard cooler is built for 1,000 MHz memory clock. Running his mental memory clock will just draw a LOT more power and create heat - without increasing your hash rate.
With the latest modded OpenCL kernels, hash rate is pretty much linearly proportional to number of stream processors and the clock they run at. So with the same card, it is ALL about core clock rate.
In this case, you've got a card that has a guarantee that it runs happily at 1,000 MHz core clock. That's with the memory overclocked too! So putting a large underclock on the memory will reduce power consumption and heat - whether it allows greater headroom with the core clock is for you to find out!
I can give some exact numbers here - with a standard 'old' Sapphire 5850 (with the 5 fat copper heatpipes), running 999 MHz core and 275 MHz memory netted me a fat 408 Mhash/s with temps below 70˚C. This card may be one of the lucky ones, because I've got 3 other 5850s and none will clock this high.
The big question is whether the additional vertical heatpipe cooler is cooling the VRMs and not just the RAM chips. If it's cooling the VRMs... then you are in luck, even if just for longevity purposes. Even when an overclocked card is running 60-70 ˚C reported temperature, the VRMs themselves often are well over 100˚C. Cool those - and the card will last much longer. You may be able to overclock the card further...
However I've found that some of my cards simply won't overclock beyond a certain speed even though the card is well within thermal tolerances. It's the luck of the draw regarding the quality of the GPU wafer.
I'd say that you'd definitely get more than 408 Mh/s simply because the seller claims it's stable at 1,000 MHz core clock, and I know what I can achieve with a 5850 at that clock!
BTW - my figures are at standard voltages. Overvolt at your peril. Though, again, if that big heatsink is protecting the VRMs, then overvolting is safer. But I'd do more research into how the *other* components on the card react to overvolting before touching those options. Remember, if all you're after is 400 Mh/s - you've got it, at core clock. No need to overvolt!
Regarding stability - just run the damn miner flat out. If you're clocking past stability, then the kernel will crash, the card will stop accepting work, the GPU core temperature will drop and the mining software will stop returning work. I just watch the temperature for an hour, and if it steadily heads to 90 degrees then it's too hot - and if it rises before falling dramatically, then it's just crashed and I need to reboot the box. DO NOT COMMIT YOUR CLOCK SETTINGS UNTIL YOU ARE *SURE* THE CARD CAN SUSTAIN THEM! Otherwise, the card may crash on boot and you'll have all sorts of fun.