I also tend to favor metal film in general, but in the Hafler interview, he referred to the “low noise” resistors being noisier than he expected. In the rest of the interview, did he explain which resistors he was talking about? I recall the PAS came with a different kind of resistor on the phono board – originally not carbon film. They were some kind of close tolerance resistor – was he talking about certain of those causing noise in the phono circuit? What did he replace the noisy resistors with at the time, I wonder?
I don't know more about the story beyond that interview. What we can conclude is that Hafler found a different supplier. Carbon-composite resistors (CCR) are a very poor technology with significant variation between suppliers, so another supplier could provide less-noisy resistors.
To understand why CCR properties vary, and indeed what that variation is, we must consider how a CCR is made: varying amounts of conductive carbon black are mixed with insulating powdered clay or other ceramic, with a powdered phenolic (the brown material used in old printed-circuit boards) binder to glue it together into a cylindrical form. The ratio of carbon (conductor) to insulator (ceramic) determines the resistance in an inverse relationship; high carbon content creates lower resistance and low carbon content creates high resistance.
The physical construction causes the CCR to act like a rectifier of sorts, must like the early CuO rectifiers and varistors, and it also has frequency-dependent effects much like capacitor reactance. All of this distorts the signal in unpredictable ways.
Again, I like metal film for the reasons you cited. However, when I rebuilt my PAS I retained the original carbon comps on the PCB and I do not recall obtrusive hiss/noisiness, at least not at listening levels with high level sources.
As per the story, Hafler simply found a better supplier. Since he was price driven, he'd likely selected those resistors based on cost. I'm not disrespecting Hafler and his crew here; they had a
product to get out the door at a specific price point, and if they missed that target they didn't have a viable product or jobs anymore.
The fact that Hafler improved his CCR supplier to the point he could ship doesn't mean that low-level noise doesn't remain, or that the circuit properly performs with out-of-spec CCRs that drift over time.
Carbon comps are okay for grid stop positions (when you have them) and should not cause a noise issue due to the low current, or so I understand.
Ummm, no, that is just audio mythology which endlessly recirculates. Time to debunk it yet again.
Carbon-composite resistors have
terrible properties. Simply awful.
One is the Voltage Coefficient of Resistance (VCR) which causes the resistance to drop with voltage, beginning at about 100 VDC and about 6% to 15% depending on vendor. If the signal amplitude is small or the voltage is constant or low (as it would be for the grid) no beneficial change occurs, just more noise and distortion. Erie did a study which showed a few percent change at about 350 VDC. The effect, however, is quite significant at several kV.
So without a large swing in voltage, and a high-amplitude signal that won't be drowned out by noise, all that happens is the CCR is a poor tolerance resistor.
CCRs are hygroscopic and the absorbed moisture permanently changes the resistance, about 3 to 10% according to the data I've seen. Which means that equipment stored in damp basements may have had the CCR values altered long before it began living in a nice climate-controlled living room.
On top of that, the CCR causes significant harmonic distortion. But since humans like even-ordered harmonics, this is considered to be a "warm" sound. Same thing happens with capacitor-bypassed cathode bias; the even harmonics are seen as an improvement and this is reported as mellowness and fullness.
The
only place where the VCR property of the CCR could possibly matter is with high voltage swings, such as with a plate resistor. Claims about mojo for any other use is not true or accurate. The guitar world is particularly vulnerable to these claims, btw.
Switching to my audiophile hat for a moment, I’ve gotten the impression that carbon comps contribute to the classic tube gear lushness/fullness/richness of sound. Metal films still allow lushness while resulting in more of a clinical, clean, distinct quality to the sound. One thing I noticed when I replaced the carbon comps with mil-spec metal films in an old Dynaco solid state amp (original Stereo 80) was that the carbon comps also seemed a bit livelier, while the metal films sounded a bit “slower.” Subjective audiophile dross? Perhaps. But, I have read similar comments from others. In any case, I’d probably tend to favor the metal films in general, but if the original CC resistors are in-tolerance and not causing a significant noise problem, I am not inclined to think it is as much of a priority to change as it would be to, for example, make sure the O.P's power supply is properly functioning.
Yeah, it's called
distortion. (Second-order and higher-order even harmonics, to be precise.) Liking distortion doesn't make you a bad person, it doesn't (always) mean you have bad taste, and it won't undermine civilization. But it
does mean that the signal is altered by the amplifier, somewhat defeating the goal of having an amplifier be a straight wire with gain.
Don't get me wrong: I love distortion in guitar amplifiers. I also don't mind or actually like certain types of distortion in HiFi (and guitar) amplifiers. I often enjoy the sound of analog processing which tinkers with or alters the signal: expanders, compressors, impact and sub-harmonic restoration, intra-aural cancellation, noise filters, etc. But I'm not in denial about what this is or why I like it. Audiophiles need to wake up and start admitting their amplifier with Magikal Mojo distorts in ways they find sonically pleasing. Yes,
distorts. Altering a signal distorts it. Hopefully for a more pleasing sound, but it is still distortion.
But I can't complain because it is all free. 
(Ok, I can still complain, but nobody will listen...)
