DEC 3000/600 Alphaserver problems

I have a rackmount DEC 3000/600 Alphaserver containing a H7816-BA power supply which doesn't work. As I mentioned before, this power supply is a real pig to work on. I also have two tabletop DEC 3000/600 Alphaservers which contain H7816-AA power supplies which are similar but not identical. The main difference seems to be that the H7816-AA power supplies contain four integrated fans (and lots of dust) while the H7816-BA relies on one large fan external to it. I am hoping that comparisons between a working H7816-AA and the failed H7816-BA will help diagnose the cause of the failure. So I dug out one of the tabletop machines, took the cover off and powered it up to ensure the power supply actually works before I go any further. The green LED came on, the fans turned and the diagnostic LEDs lit up. After running for less than a minute, there was a cracking/popping noise and a small spark visible from somewhere around three unmarked orange surface mount components on the I/O board (not in the power supply) plus a whiff of cooked electronics. The power supply kept running and the diagnostic LEDs remained lit but I powered off quickly and poked around with my finger to try to find a hot component without success. I examined the area with a magnifying glass under a strong light but I was unable to find anything damaged. I wonder would these orange components be tantalum supply decoupling capacitors? They seem to have a small pip in the solder at the positive ends. After powering back on, two of them had 5.0V across them and there was 4.87V across the third slightly smaller one, there were no further fireworks (so far anyway) and the SROM mini-console works. I have had similar experiences with one or both of my two tabletop units in the past and I was unable to track down the culprits then either. Neither unit works properly due to cache failure issues. I wonder could it be plausable that the caches are not functioning correctly due to prior failure of several supply decoupling capacitors on the system board? Unless these devices have markings on the underside, I have no idea what their capacitance values are. There is probably no point in trying to measure them in-circuit if there are several all over the board in parallel with each other. Unsoldering them and trying to measure them then isn't going to work either if they have failed. Any suggestions for something to replace them with that is less likely to go bang in the future? Could I get away with using components with leads to replace them instead of surface mounted ones or would that introduce too much inductance? Regards, Peter Coghlan.