For the specifications below, what architecture/topologies would you use?

4 x 35 V output all isolated, 5% regulation Slow turn on after 30 second delay. Overcurrent protection needed. 400 W total power on these outputs

+/-60 V output 1% regulation ground reference. Must come up first. 5 W total power

12 V Output. 30 W total power ground reference. Comes up with 60 V output. 5% regulation.

Control bias as needed.

120 - 240 VAC input

PFC optional.

Performance far more important than price. High density and low-noise a plus.

If using something like Vicor building blocks, how would you incorporate sequencing and protection? What are the "gotchas" with their parts?

If building discrete power, what would your preferred sequencing and protection be?

Hmm... I think I would divide it into 2 separate converters: One to make the internal supplies, the +/-60 V output and the 12V. Could be Flyback followed by a linear regulator in order to cope with the high precision on the +/-60 V output and flyback to keep the converter small.
For the main output I would suggest full bridge to get the maximum out of the magnetics. Current limit on the converter to cope with the overcurrent protection.
The squencing will be inherent since the flyback will supply the full bridge. The 30 seconds can be done with a RC filter on the softstart pin that is on most control ICs.

That is a tough one! Efficiency is going to be tough as well. Multiple converters and mixing of topologies is a definite. Most would turn to bricks. Seven bricks total and they might have to be custom based on power and overcurrent. Vicor is probably the best, but many other brands leave much to be desired as to performance, and very poor support.
Had a similar situation back in the early 80's and the chief engineer came up with interleaving the converters with multiple forward. Lots of parts and loops.
PFC might be a plus to give a regulated bus to start with. My first thought for the 4X35V would be having a bang bang front end putting out an isolated 4X40V or so and then have 4 buck controllers. Maybe the 12V might be included.
The +/-60V @ 1% is the real tough one. Separate converters for each for the 1% regulation. Parts count will be high and for density the smallest possible components will be necessary, but the magnetics will still be a large volume of space.

Thanks Ray for the 'stump the chump' discussion. Looking forward to all the other approaches.

I agree with Lasse without PFC, but with a PFC, we get a well regulated input voltage, so I would put a LLC resonant converter for 4x35V output instead of full bridge because firstly the secondary rectifiers will not suffer from high voltage spark, secondly the cross regulation of 4 separated 35V would be much than full bridge and thirdly the components, primary MOSFETs and drivers, would be half of full bridge.

I second Lasse's approach. This is how I did it on a recent design. Logic and a negative supply comes up first, then 12V which feeds all other conversion stages. High voltage is a separate converter. If efficiency is important also for the 12V you need a flyback with a recuperating winding for the primary instead of just flaring off.

As for commercial blocks I never use those. Without naming the mfg in public, but it is a big famous one: They told my client to never hang an incandescent test load onto their module because that can supposedly cause it to become damaged. That completely eroded my trust in them. All my switcher designs, and it's dozens by now, can tolerance negative impedance loads.

What surprises me is "PFC optional". Even on 400Hz aircraft systems that is usually no longer the case. Which market is this for?

Question about the requirements: Is tracking required on the +/- 60 V outputs?

Looks like audio amp power supply - LLC or similar converter. Noise would be the main problem.

The 4X35V, each isolated with 5% regulation seems to me to be the tough as well as the 1% on the 60V outputs. I'd like to hear more specifics as to these requirements.

Joerg: do you think the comment on blocks was an engineer being cautious, or a real concern? I haven't heard this issue before.

60 V don't need to track as they come up, if that is what you are asking, Bob.

Those of you who suggested building power stages - why not use Vicor or other company building blocks?

And a follow-on question - if you design your own custom power supply, is this more reliable than configuring Vicor modules? We would be talking about maybe 7 or 8 modules to do this power supply, and have to deal with sequencing, current-limiting, shorts, etc.

What have your experiences been in this area?

I would suggest also a resonant converter if low noise is an issue. For the four isolated 35V outputs you could use a series resistors with a parallel relay, after the large output capacitors have reached the nominal voltage the relay will short the resistors. This gives the slow voltage ramp with safe operating for the power supply and caps. For the 60V output I suggest linear regulators to be able to have the 1% tolerance considering the low power of 5W. The loop can be closed on the 12V line and should be able to regulate within the 5% tol.
For 400W power and a resonant converter topology the active PFC I would guess is a must.
For protection I would implement two triggers: Over-current monitoring in the primary (some controllers have opamps and disable for this) and over-voltage in the secondary also to disable the controller.
If the power supply should have some special requirements regarding voltages, currents, layout, chassis placement, noise, protection a custom design should have a better performance in the end.