Receivers with very narrow bandwidth PLL fully synchronous demodulation might work with a single sideband vision transmission with pilot carrier, but even then, some care would be needed with the IF passband to obtain on the one hand, a sharp cutoff on the non-sideband side of the carrier, whilst on the other preserving the amplitude and phase linearity down to virtually DC on the sideband side of the carrier. It is not like voice-quality HF SSB reception, where audio frequencies below around 300 Hz may be ignored. With HF SSB broadcast relays, 100 Hz was the customary lower limit (90 Hz for the BBC), as compared with around 50 Hz for normal broadcast practice, and I suspect that some phase errors in that vicinity would have gone unnoticed. And very narrow bandwidth PLLs for vision demodulation have the disadvantage that any incidental phase modulation is carried over to baseband.
That said, whether SSB transmission with preservation of amplitude and phase linearity would be feasible is another question. My understanding is that the original choice of VSB rather than SSB had as much to do with the transmission end as the receiving end. VSB allowed the transmitter sideband filter to be reasonably removed from the carrier frequency.
PLL fully synchronous vision demodulation became more common in the later analogue era, but by no means universal. The probably more common quasi-synchronous (QS) variety, as conventionally implemented, might not have been very comfortable with SSB signals, the more so those with reduced carriers. Any sideband asymmetry in the signal presented to the carrier channel limiter results in demodulation errors (distortion), a facets that is seldom noted in the literature. The IF passband Nyquist slope thus in and of itself causes errors in conventional QS demodulation. And the errors would be worse for a fully SSB signal with zero sideband on one side. With a VSB signal, one could extract the vision carrier using a narrow bandwidth (i.e. less than the twice the VSB bandwidth) SAWF ahead of the main SAWF (and I understand that this was done in some BBC receivers), but that would not work for SSB, unless the filter could have say a 50 Hz or lower bandwidth. It may also be seen that the Nyquist error will be reduced with a gentler Nyquist slope, so that receivers using QS demodulation would ideally have an IF bandpass that fully utilized the transmitted VSB, even though with SAWFs, a steeper-than-needed Nyquist slope was feasible without compromising phase linearity.
As an aside, the Nyquist slope error was responsible for some of the intercarrier sound buzz with both rectifying and QS vision demodulators, and from that source it could not be removed by hard limiting in the sound IF channel. (There was though a work around back in the rectifying demodulator days, e.g. see the B&O 3000 circuit.) Nyquist slope buzz became intolerable in the (modern) stereo and multichannel sound era, hence QSS demodulation for the sound channel. (I say the modern era, because the French had bilingual sound transmissions with System E in Algeria in the second half of the 1950s.)
Regarding System I, by design I think it made the best use of an 8 MHz channel when following the established wisdom re guard bands; one assumes that the use of established guard bands was taken as a given in the investigative work. The French “cribbed” an extra 0.5 MHz by overlapping the guard bands. Likely this was based upon their experience with System E channelling. Whether negative vision modulation was a better choice for System I is I think very debatable, and probably a topic where if one asked 12 experts, one might receive 13 answers. In that era though, the body of opinion seemed to favour negative, so System I was ‘going with the flow” as it were. System L was counter-current simply to make easier the design of dual-standard receivers for Systems E and L.
On the bandwidth issue, here is an interesting quote from Carnt & Townsend, Volume I, from Chapter 4, “Transmitter Coding”:
“For 625 lines the suggested chrominance bandwidths are the same as for the American system. If an 8 Mc/s channel is available for each station, as in the U.S.S.R., the optimum utilisation of the extra megacycle of bandwidth is probably to increase the luminance pass-band by 0.5 megacycle and use the other 0.5 megacycle to increase the vestigial sideband of the vision carrier to 1.25 Mc/s. The latter reduces the effects of vestigial sideband distortion, particularly on negative modulation. If the sub-carrier frequency is left unaltered, the extra luminance bandwidth can be used to increase the chrominance bandwidth and to ease the problem of providing a sharp luminance and chrominance cutoff at the sound channel frequency.”
Looking at the colour side of things, the chrominance subcarrier needed to be of as high a frequency as possible to minimize its visibility in the luminance channel, and with System M and B the tradeoff had been sideband asymmetry in the chrominance channel. But the subcarrier frequency chosen for System B was evidently sees as high enough to be satisfactory from the (in)visibility viewpoint, such that the additional vision bandwidth available with System I could be allocated to avoiding significant chrominance sideband asymmetry.
Thus one might say that System I was better optimized for future colour transmissions than was System B. On the other hand, System N was going in the other direction, as it would have required a lower, and thus one assumes more visible subcarrier frequency, and with no relief from sideband asymmetry.
Had “quantity” won the argument over “quality”, with a modified (positive/AM) version of System N being chosen for the UK circa 1960, one wonders whether inverted channels (vision above sound) would also have been chosen to align with System A and so minimize dual-standard receiver complexity. Receiver simplicity would have argued argue for a common sound IF (38.15 MHz) and thus a vision IF of 33.65 MHz for 625 lines, which one hopes would have been a satisfactory choice, although at least at UHF, channel assignments could have been planned around it. Also would have been the question as to whether to use the European 625-line waveform (5 equalizing pulses) or that of System N (6 equalizing pulses.)