One of my goals as a new ham is to homebrew my own transmitter (using vacuum tubes, aka valves). Initially I'll stick with CW for the transmitter, though over time I might build an AM modulator or even attempt an SSB modulator/filter. For the CW iteration, I plan to use a vacuum tube Hartley oscillator, with the tap on the oscillator coil positioned to minimize drift on key-down.
I've read of some CW transmitters with high voltage at the key, suggesting they were interrupting the B+ to the output tubes (for power levels above QRP, this may be several hundred volts; for higher power, it may run to several thousand). Obviously, if I do this, I'd use a relay to keep the high voltage inside the case.
For other designs, however, I've seen reference to keying the oscillator -- either interrupting the output of the oscillator before it reaches the first output tube's grid, or interrupting the plate or grid on the oscillator's own tube to completely stop the RF (potentially also leading to relatively high voltage at the key).
As I understand it, both can work well enough to have survived until transistors replaced vacuum tubes for most applications, and obviously the quality of the keyed output depends on multiple other factors (stability of the oscillator, rise and fall time of the chosen keying control, etc.).
From the standpoint of signal quality, however, is there a good reason to prefer one over the other?