I've built quite a few amps, and I hear no audible difference between hexfreds or 1n4007. In fact, I'm pretty sure there is none, and I'd bet money nobody can tell in a double blind test. That's why Mesa uses 1n4007s even in its "high end" amps such as Mark V and Recto in all its many variations. If there was "hash" because of this, the likes of John Petrucci would have told them to remove it.
You also can't use "less filtering" no matter which diodes you use. Most of that filtering capacitance filters out 60Hz hum. For high frequency content there's usually a 0.1uF film cap in parallel with electrolytics to filter that out as well, at least in well designed amps. Mesa uses 0.047uF, about half that. The reason why "less filtering" might be desirable is because it creates greater voltage sag in a cranked amp. This is an accidental "feature" of older Marshalls where ol' Jim wanted to save a buck (or a quid as the case might be) and didn't install enough capacitance in the power supply.
And yes, 5U4 recrifiers do sag, even in parallel. Per the spec sheet, each tube drops 50V at 275ma, which yields resistance of about 182 ohms. This will vary with current slightly, but not as much as to be audible even with the golden-est of ears. So 90 ohms for two in parallel then. You could even make sag adjustable for about $2 by adding a simple MOSFET-based sag circuit if you use diodes.
BTW, the "variac" mode in Mesa amps means quite literally what it says on the tin - it is accomplished by way of having a separate tap on the mains side of the power transformer (the "primary"). The entire amp save for the switching circuits (which are voltage-stabilized) is in effect under-volted in this mode. This does affect the tube rectifier somewhat, of course, its plate resistance goes up due to not enough heater voltage, but that too could easily be accounted for with a different resistance in series with the diodes, or the aforementioned MOSFET sag technique.
The goal here isn't to "blow you up", the goal is to share knowledge.
