Now: in most cases there are 
  - 3 mini resistors 1/8 Watts 1% (R41, R52, R55)
  - 1 large resistor 5 Watts (R19)
  - 1 zener diode (VR9)
 - 1 capacitor (err ... must be the 220 microfarad elco at C46)
 - BUX85 (Q10)
 - maybe the BTA06 (Q5) blown.
 While there: check 1N4007 (D16) as well.
     All these parts belong to the "standby circuit" that keeps the +5V sensing voltage up as long as the PSU is connected to AC mains. Obviously the parts are underdimensioned and get toasted over the years. When the first resistor goes south it takes the other components with him ... and sometimes the large Hybrid chip below the coils as well, which is the over / undervoltage detection circuitry as far as I know. 
     The above mentioned parts can be obtained in good electronic stores ... the hybrid module cannot. It is a special part. "Issue 4" and above models contain an additional 1N4007 Si-Diode in series with the resistors. 
    A similar malfunction is known from several Apple power supplies ... accidently made by Magnetek too ...
   The *cooling design* was okay - but the italian manufacturer loused up the inards by using inappropriate parts in the 35 / 56 / 76 PSU. 
   Same on some Apple power-supplies: there is one electrolytic condensor specified for lower temperature ranges. When this thing dies after some years it takes some resistors, a diode and a VMOS transistor with it. Result: dead PSU.
   The UK-made Plessey PSUs are not known to blow up. Sadly these are rare. 

   Recommendation: try to get some spare PSUs. The IBM models 35, 56 and 76 small desktops use the same PSU (and Mod. 40, 57 and 77 large desktops use the same larger PSU as well)
   And: use a common line breaker to fully separate the machine from the AC-power when not in use for a longer period of time.

>Thanks for the info.  It appears that the PSU is made by Delta.  The IBM P/N is 79F3367 and the FRU is 79F3443.  So hopefully it isn't he PSU if you say that the Delta PSU's aren't known to have a signifigantly high failure rate.

The FRU is 79F3443 for all sorts of "Mod. 35/56/76" 118 Watt power supplies, the P/N might differ. The objection on the failure rate however is my own - and derived from a decade with these units from which 90% had the italian Magnetek PSU - and all of them were Greenock/UK productions (serial number starts with 55-). So the statistical number might be invalid, because I did not have too many units under my hands with Non-Magnetek PSUs. But these *are* know for significant high long-term failure rates.

While being on it: I have seen a few 9556 with ASTEC / Taiwan PSUs.

>As for the delay after power off, I was aware of that, but this problem makes the machine unusalbe for a couple of hours afterwards. [snipped] Then if I come back and power it on a couple of hours later, it runs like normal for only a few minutes.

Hmmm. That sounds a lot after a shot / dying PSU. But can be the planar as well. Maybe it is a coincidence, but I have a Mod. 95 with a Delta-PSU. Once you switch it on it takes about 10 - 15 seconds before any light comes on and the fans start to turn.

>Out of curiosity, if the PSU was failing, would it get hot or anything like that?

Not neccessarily. On the Magnetek-units the failure is basically caused by a too high thermal loss on one or two resistors, which are dimensioned for a loss of 1/8th Watt ... over the years they are "cooked" - get brittle, until they are so much worn that they burn off. A short "pop" - and they're gone. Now the standby-part of the PSU is non-functional (does not supply the sense voltage
for the electronic switch any longer) and the main power supply cannot be turned on. Often the failing resistor causes an overvoltage in a Z-Diode - which damages a Thyristor ... which fries some other components ... a little firework. The PSUs used to have a certain "electronic smell" after burned resistors and capacitors. When you disassemble the PSU you will found some parts that turned into coal and a printboard having turned to a dark brownish color in some areas.

Sometimes the PSU still works - partially. You can turn it on with unplugging / plugging back the AC-plug ... but the on/off switch is no longer functional.

I think yours might be partially damaged, so that the "Power Good" line from the PSU to the planar hangs in "undefined" condition, which holds the Reset line of the CPU low. Therefore the HDs and fans start to spin - but the board does not start up the power on self test (POST) and memory count during that. This -however- could be a symptom of a shot board with a thermal malfunction as well.

Therefore my recommendation: try to get a spare PSU to test the unit with it. It is good to have a spare anyway. Maybe there is someone out who can lend out one for a test ... ?
 
 

IBM PS/2 - 9576i Will Not Power Up
  Power Supply internal component can fail after short period of use.
Problem Isolation Aids:
Examine power supply.
Visual Inspection:
With cover on, Magnetek Power supply has Air Vents at the top right hand corner which can be seen through the right hand system lovres. With the System cover off, there is a Vendor Label on the fan openside of the power supply clearly displaying the vendor name.
DETAILS:
POWER SUPPLY VENDOR - MAGNETEK 118 WATT
IBM P/N 79F3367 (FRU P/N 79F3443) HEADER CODE B1JT1
Fix:
REPLACE WITH FRU - IBM P/N 79F3443

1.Check for continuity between pins 4 and 6 (switch) on P2. 
2.If the switch is OK, short pin 4 to pin 6 on P2 and check the power supply voltages.