Hi Andy,
If you'd applied what would conventionally be referred to as 10dB of feedback, for the same drive voltage at the input, the output voltage from your amp would fall to 1/3.16 of what it was before. the power would fall to one tenth! the distortion would fall to 1/3.16 in theory due to the feedback, but would fall further due to the reduction in output level.
So an amplifier designed to have a negative feedback loop is designed with the appropriate excess of gain. run without the feedback connected it will have stupid sensitivity. With the feedback operating, the gain falls and the sensitivity comes down to what was planned.
Getting the higher open-loop gain means pushing the input and phase splitter stages for more gain. Which means higher resistance anode loads, higher Gm valves etc. These things tend to mean more phase shift which ruins stability, and so very careful design is needed to get the wanted gain while minimising phase lag, and not letting the measures which boost the gain make more distortion.
So an amplifier which will do the right job with feedback operating will show too much gain and too great sensitivity when run without it.
Designing a valve amp which works moderately well open loop is pretty easy. Compared to transistors, valves will let you get away with murder. Designing a valve amplifier which gets all the benefit it can from a closed loop is a serious undertaking. So most people building their own stick to a reference design like the Mullard ones which don't push things too far and will work stably with fairly mundane transformer designs. Many companies' amplifiers are nothing more than light variations on these. Some companies, Leak, Quad, Radford pushed things further, pretty much to the limit, needing fancy output transformer design to make it work.
It can be taken a lot further, but you have to ditch the output transformer and ditch the familiar architecture. I did it once out of curiosity. The amplifier was huge, power hungry and of a complexity where I might as well have been using transistors. It could have appealed to some people who judge on cost and looks... reassuringly extreme on both counts.
In traditional education, the maths involved in feedback analysis isn't taught at O or A level, though the foundations are. It is a major component in the maths taught as part of an engineering degree. Most people find that scary, and engineering degrees have a general reputation of being all mathematical and hard. However, outside the traditional education system, in the real world, it is quite do-able if you want and are interested. It isn't even hard. Some bits like algebra need a fair bit of practice until you get familiar with them and you start to see patterns. Other bits involve weird concepts like imaginary numbers for frequency and you have to suspend your disbelief until you see them doing useful things to designs in our universe!
I've seen people who dropped out of sciences without O levels get their heads around this stuff. It rewards them with a toolbox which can do all sorts of things and the feeling when seemingly disparate things snap into place is wonderful. If you have the curiosity, and are exposed to these things you'll absorb it slowly and it will sneak up on you. It's a lot easier to teach these things to someone who hasn't been told that it's supposed to be hard, which tells you something
David