The Inuit indians are supposed to have 23 (or is it 46?) different words for “snow”.
Similarly we have a number of words that mean the same thing with different twists - ground, earth, common, chassis, deck, return - depending on the context.
Attention gets paid to the HT distribution side and bypassing but the mirror-image return side is normally implicit in circuits with little ground symbols 
scattered all over the place.
This gives the impression, reinforced by a solid plated steel chassis, that ground is everywhere; signal returns, power returns, just wang 'em to chassis wherever you happen to be, she'll be right.
Well quite a few amp builders did just that, assuming the chassis to be equipotential (the same voltage) everywhere.
This is nice and simple and certainly it works. Well mostly. But mostly not well.
The main difficulty is that letting the earth returns take care of themselves you wind up at their mercy. Not having planned and managed the return paths with as much care as the active paths, if something goes wrong, the amp is unstable or hums badly, you really have no idea where to start because you haven't given it any thought yet.
If you think grounding in a problem in an audio amp, spare a thought for the designer of state logic systems. Here grounding and the definition of system earth is such a problem that it is normal to consider and distribute power and return paths as transmission lines and with as much care as the signal paths, generally trying for a mirror-image of 'go' and 'return'.
It's easy to see that as signals do things to the stages they will also be impressed on the power supply rails, hence the need for de-coupling in the HT. But what about the other rail, chassis/ground?
These signals are equally impressed on the return rail. Therefore the best place for a stage bypass cap is right across the stage itself (remembering the bypass cap is a short circuit to AC signals) with its feed resistor right nearby, but not so close it will heat the electro.
Many reasonable stereo amps have thick ground traces leaving a single point, generally the -ve end of the last power supply filter cap, and running to various stages isolated from each other, even when a bunch of them run together right across the board. Looks stupid until you understand why.
In this case what they are doing is getting rid of common resistance (impedance) in the return paths. A bit of return resistance isn't a problem until something else starts developing a voltage across it and injects its signal into that stage.
So the first problem to avoid is shared resistances or resistance in common.
It was quite normal to ground your signal screens to the chassis over here. And over there. And sometimes there and there as well.
In this case earth loops are formed consisting of the chassis and screen connected in parallel. Add some stray AC field from the heater wiring and significant circulating current can flow around this shorted-turn loop. If cable screening were perfect all this current would flow on the outside and this wouldn't worry us, but it isn't and doesn't, and so this circulating current induces a signal in the screened conductor and we have an insidious hum or the like.
This is the dreaded earth or ground loop problem, but within the equipment. These should be designed out, but where it has already been designed in one treatment is to insert a 100 ohm resistor in series with the screen at one end. This generally has little or no effect on the interconnected stages, but reduces the circulating current by orders of magnitude, 1000 times if the loop resistance was 100 milliohms.
The base cause is 'economising' by using the screening as signal return as well, and there are wirings and schemes that avoid that, balanced being the most common.
So the second problem to avoid is forming low resistance loops.
Strauss amps commonly had a signal ground consisting of a tree of heavy guage tinned copper wire running the length of the signal path and grounded to the chassis at only one point (in the power supply as I recall). They were also among the few to use input jacks isolated from the chassis.
But most didn't. And this presents us with an interesting problem. Many of these amps could be improved, that is quietened, by a bit of grounding design they missed out on.
If you were building these days you would use insulated jacks, single-point earthing for each stage and a fan of grounds back to the main single-point chassis connection, similar to above.
Retro-fitting a guitar amp is a bit more of a problem, particularly as it may emerge the best place for the earthing point is the grounded input jack.
Well Pat Hawker had a look at this sort of problem in Ham AM modulators (possibly the closest living relative of the guitar amp) and came up with a scheme that should suit our needs very well.
It sticks with the line of development above where you come back to a single earth point in the power supply, but with a twist. Instead of connecting to the chassis there, you run a heavy wire to the grounded input sockets on the front panel and call that your chassis earth point. Note the first stage runs there direct, and may be useful with the EQ stage as well.
This scheme would seem to be very suitable as a retrofit on a hummy head if grounding seems to be the issue and other common causes (e.g. heater wiring) have been eliminated, or where you are getting stability problems from unintended feedback.
