Tired CRT unfortunately. Especially ones that use automatic kine adjustment poorly designed ages the cathodes emitters fast. This feature is undesirable on Sony and some models including Sharp. We had a Sharp TV for short time. CRT went downhill in space of 2 years and it had that feature.

There is nothing to help with the filtering as it is impossible due to dynamic nature of video signal through the CRT gates. The filament heater is thick and doesn't show any AC due to cathodes heated to orange heat and gates is what introduces ac component into electrons flow to the shadow mask creating video image. HV is filtered by the CRT's capacitance.

Most noticeable is red. Blooms brighter than usual, losing fine details, and tailing comets. Worse with brightness when increased.

Cheers,

That is not called bloom but breathing instead. Bloom is spot size change with brightess (higher brightness = lower local focus) while breathing is image geometry change with brightness.

Breathing is caused by sag of HV due to relatively high impedance of the winding, and the effects of it are normally countered through EW correction. Normally beam current is measured at the lower side of anode winding of the flyback ("ABL" is the common term for signal that carries this measurement) and image width is adjusted proportionally through EW correction to counter raster expansion (lower HV = electrons move slower and spend more time in deflection field and thus deflect more, leading to expanded raster) from the sag on the HV with image brightness.

From what I see on TVs, the matching is not very good and you can see very large variations in geometry. You can mess with the time constants and level on the ABL signal and get improvements but you'll never get solid geometry like a monitor would produce, you really need regulated HV for that or very least a higher quality flyback with lower impedance anode winding.

For example my 32" Grundig got much worse geometry performance after I had to replace its flyback earlier this year, the replacement simply is shit compared to the original and while I could get improvements by doing previously mentioned adjustments, I could never reach the original performance. Ideally one decouples line delfection from HV generation to get regulated HV, like in a monitor but this basically requires redoing the entire TV...

I'm using the CRT Repair FAQ's definition of blooming:

Blooming is defined as an expansion of the raster or horizontal sections of the raster with bright material. For example, switching between dark and light picture causes the size of the picture to expand by 10%. A slight change in size is unavoidable but if it is greater than 1 or 2 percent from a totally black image to a full white one, this is either an indication of a defective TV or one that is badly designed. The cause is poor low or high voltage regulation.

What I'm confused about is why the blooming is so bad, but the TV otherwise works fine. Picture is bright and sharp as a tack with a component source. All the major power rails are sourced from one big power transformer fed by a SMPS circuit, so my next instinct (the SMPS controller IC or its buffering caps are bad) would mean the whole TV would be dead or malfunctioning, right? This was a high-end TV when new, so brushing it off as "TVs are just bad" doesn't ring true to me. I have a much cheaper (albeit smaller at 20") CRT TV which has the expected 1-2% blooming.

Ok, interesting about that definition.

The problem is definitely because of the flyback's inability to maintain HV regulation, effort on the main PSU end won't yield much results unless you actually had bad capacitors and B+ itself sagged. There can be minor improvement by having more capacitance on the B+ rail but it mainly improves jitter performance and not so much geometry.

The HV winding wire in the flyback is still about 0.05mm thick with roughly 4000...5000 turns even if the flyback is for a 40" or a 14" TV, but a 40" TV will have much greater power demand and with it the output will sag much more and cause proportionally higher geometry shift, even when the flyback has internal smoothing capacitor as it is still only a temporary reservoir and when you get a white raster things are going to sag and stay sagged. Sag is proportional to beam current, shaped by capacitance of the kinescope itself since it acts as a smoothing capacitor. I have measured about 3000pF on 32 and 36" kinescopes.

My experience with the larger TVs has been quite consistent, they don't have particularly good geometry performance (2cm shift on both sides is quite normal) unless they're actually high end, meaning they decouple HV generation from line deflection so that they can have actual regulated HV but I have only ever seen that in (presentation)monitors.

When line deflection is also repsonsible for HV generation (as it is with majority of TVs) it is impossible to regulate the HV since it disallows modulating switching time of flyback and also disallows modulating B+ itself, both which will result in big change in line deflection itself (i.e increase of B+ will both widen raster and also boosts all the outputs on flyback, including filament voltage and similar effects happen when flyback primary gets PWM'd too). Only thing possible is to compensate for the geometry shift through EW correction circuit (diode modulator in majority of cases) using beam current monitoring signal from flyback and there's often room for better matching of the time constants etc. in the signal processing steps prior to it entering in one of the big chips from which all the deflection drive signals come out of. Ideally you have a better flyback with lower impedance HV winding but that's one thing you're unlikely to get... Compatible/replacement flybacks are often worse too...

I will look in the schematics now

EDIT:
* That TV has a traditional combined Line+HV circuit along with regular diode modulator based EW correction circuit. Nothing "high end" as far as electronics itself goes.
* You can try to adjust ABL Gain and Start Point (V16, V15) in the service menu, perhaps it will give some improvement.
* C451, 452, 522, 523 and C617 will delay and shape the ABL signal, these parts should definitely be checked if they're in spec. They are there to try to shape the response of actual flyback's output sag with beam current derived info and there is some playroom in there. If there is bad matching the correction will over or undershoot during fast dark/bright transitions. If there's noticable static difference in geometry between a dark and a bright scene there's some level matching needed too, which is going to involve the resistors, R462 is a good canditate to adjust. Pin28 of TB1253 is the ABL signal that is used to modulate EW correction and frame drive to compensate for the these HV sag caused geometry shifts.

Without desoldering everything (requires disconnecting all the cables that plug into the main PCB, then prying off the neck board and propping the whole thing up in an odd position to minimize tension on the flyback wires) I was able to test R462. It's a microscopic SMD resistor on the PCB underside, and after reflowing the solder to get a good reading it tests at 20.4kOhm, barely within 10% tolerance. What do you mean by a "static difference"? I can play video games with blinking dark/light areas and see vertical lines bow in and out as the line brightness changes. I don't see noticeable stretching at light-dark transitions in a static image, ie. faces and other shapes still look correct within horizontal bands that have about the same average brightness.

Static difference means a change that stays in place between white vs dark/grey raster, i.e bright raster stays 1cm different from a dark one. You can use "240p test suite" to test bright vs dark image and see how edge of raster changes.
For example my 32" Grundig overshot the compensation, white raster shrunk compared to dark one and I had to adjust a value in its service menu to get things equal and adjust the equivalent capacitors to stop it from over/undershooting during fast transitions (i.e whole frame rapid bright and dark transitions). These sort of tests should be done with full screen changes, once those are sorted you can try to to deal with changes between say bright section vs dark one which absolutely requires adjustment of those capacitors to shape the freq response on ABL signal to match the actual HV level as closely as possible.
The capacitors are the most important parts here, and on that particular TV you have they're mostly e-caps and subject to age related effects. In case of my TVs, only film caps were used and they're not subject to aging like this.

"breathing" is picture size changing with brightness. Blooming, certain color losing sharpness becoming blurry blobs, comet tails, sparkling, raster lines appearing is blooming or tired CRT. The article writer might mistaken that.

I worked in TV repair shop before.

The breathing is due to flyback too cheap. There's proof; We had 3 types of flybacks for one type of RCA chassis. Original, junk clone and high quality copy (Diemens). Both worked correctly but the junk clone breathed so bad that you can see the top of flickering bars scanlines appearing when picture shrank too much.

The B+ supply regulation is very solid in most cases and there is second low current, low voltage feedback output is from tapped into the HV circuit via a internal resistor near the ground that is fed back into the main IC, aka "Jungle IC", that actively compensated for the minor breathing via adjusting another part of circuit that has to do with HV.
The horizontal pulse drive driving horizontal transistor's gate is constant frequency and cannot be altered otherwise. The width correction is done at the horizontal circuit, the other half called pin cushion capacitor and pin cushion diode is in parallel with fixed horizontal capacitor with diode, the pin cushion adjustment is vertical modulated onto the horizontal pulses via a inductor into the pincushion circuit. This only fixes the horizontal width and pincushion corrections. Most CRT stuff except monitor has basic active vertical height correction, but for active advanced vertical pincushion correction is rarely implemented. High end TV and quality monitors has both.

Cheers,

Bumping this since I finally replaced the capacitors. It took this long because I also managed to pick up the next model year of the same TV (27F640) off Craigslist, hoping one would be in better shape and I could sell off the other to a a less-nitpicky home. This was lucky because I can run the two side-by-side and feed them patterns with a splitter. After replacing the caps on the '631, it has the same amount of bloom as the '640. The '640 was kept in a climate-controlled bedroom and rarely used, so the caps are less likely to be dead than the older one which was sitting in a garage for who knows how long. Thanks to everyone who responded, you just can't find reliable info about how this stuff actually works on the Internet anymore. Now I can pick one (or none, having a 100lb TV has become somewhat less appealing after having to live with one on my workbench for months) to keep and offer the other one up as restored.