This is purely constructive feedback / design discussion, on the sealing of viewports, as I don’t have any knowledge how much design experience you have with this and wouldn’t want to be presumptive please take this as no insult. I have no doubts that this design could work, and is at least theoretically sound. There are some aspects that concern me, from my experience.
You start by mentioning dimensional tolerance, while the dimensional tolerance in seals is important, of equal importance is the surface finish quality. Luckily this is a circular face seal, which with the proper trepanning bit finish quality would normally not be a major concern, however with this part as the first groove is at the bottom of that deep circular boar that may cause some issues, chatter from a long tool, and perhaps clearance. Overcomable issues most likely, but does not look fun to try, and hard to evaluate for yourself I imagine.
The free floating lens sandwiched between two face seals also gives me pause for concern. I guess with this size the likelihood of a leak caused by torsional displacement would be very low, we have typically 2-3" diameter viewports where this can be an issue. However face seals like a lot, a lot, of pressure. Seen face seal covers need torqued down so hard they bend the pressure rated metal covers. Maintaining/creating enough pressure to get consistent seal pressing between two o’rings seems like it would be a concern. I guess physics suggests that this should not be the case via newton’s 3rd, so you could make a case that it is a valid sealing configuration. Just never thought about it/seen it before and it gives me some concerns, might have to put that on the list to test. (Having gotten to this point I considered removing this concern but felt that it was worthy of discussion, even if it makes me look like a dolt.)
This may not be exactly applicable to this design, but we seal our view ports like this currently. We have been experimenting with different sealing compounds, but we find this works well.
So this does mean that the lense is permanently sealed into place, but it provides some advantages to the traditional face seal. One the seal is not reliant on face seals which require pristine conditions on both housing and lens. Let’s say one of the times you are opening the enclosure, or servicing the o-rings, the lens gets dropped and scratched, wouldn’t want that to ruin electronics on the next dive. Next is clamping force, face seals require immense pressure to remain sealed, this pressure can put strain on the lens. This method uses the oring as a supplementary seal and largely to keep the potting compound in place during assembly; the potting compound is doing the sealing. Wherever possible we prefer to use permanent sealing like this, it is much more stable especially with glass/acrylic/polycarb, (in order of increasing need).
Again a lot of this comes from being used to a slightly different scale, but the viewport problem has been kicking in my head for some time and I am taking advantage of this opportunity to get my point of view/solution out and hear/see others.