Few suggestions, if I may:
No, I have not used center-tapped chokes. Yes, the choke is a rather hefty (thus expensive!) component. I am wondering though regarding two chokes plus two sets of filter caps, vs. 'combining' the lot and using a single choke plus double the value of caps. Economically - maybe you must do the arithmetic. Not sure here about which topology to choose, will be interested in comments from others. (I am in the same boat with the 100 + 100W 'Hallé' monster - I use a single choke plus caps, but at present not saying that is definitely better than two separate sets of filtering.)
Then another matter just for the record (since you are using a triode input rather that the EF86 - but my comment points to a general mistake made in several valve amplifiers in the past): By passing the screen to common (earth) complicates matters - simple pentode operation requires the screen to be bypassed to cathode. It might not matter greatly in a straight circuit, but in a feedback amplifier, the equivalent diagram will show that you have the fed-back signal voltage as a voltage source in the screen circuit, in anti-phase to the input. (It resembles a kind of UL topology, though different because there is no inductive coupling here.) But simulating will show a supposed pentode stage not very far from in fact triode gain, with the fed-back signal plus its distortion now included in the screen circuit, giving a rather queer state of affairs. I have asked around on other forums and nobody (including writer Morgan Jones) could come up with a reason for earthing the screen in this way, except lack of understanding by the designer.
This OT because you are not using a pentode input, but thanks for allowing this comment since you show such a circuit!
Another (technical again!) point: Again in the input circuit as shown, rather put the bias components (1,5K and 100µF in this example) inside the feedback loop i.e. between the cathode and the feedback point, not at the earth side. (Reason: Outside the NFB circuit as here, the impedance of the 100µF cap will become significant at some low frequency, increasing feedback with possible ill effect, although this will depend on coupling caps. Within the loop as I suggest, it will improve stability by decreasing first stage gain and thus loop gain at some low f-point, apart from phase considerations.)
Lastly I always suggest a resistor shunting the loudspeaker output, say something like 820 to 1000 ohm, to keep some sort of load on the tubes and improve stability should the loudspeaker load become open for whatever reason.
Thanks for tolerating me - happy building! (And watch out for that 525V! I had 600V between hands some months ago, fortunate to still be with you! Muscles ached for a several days from the jolt ....)