So we've had a lot of discussion over this in chat recently, and it comes down to this: All possible options have advantages and disadvantages. The final decision will depend on which of these we feel strongly about, and which we're willing to let pass.
16-wide terrain between steel: This is rare in currently existing levels. (The level I cited, Cold Irons Bound, is one we've decided to remove -- for reasons completely unrelated to the digger/steel issue; it just happened to be the weakest of a group of levels we were deciding between.) Nevertheless, I think it's important to consider what possibilities we are giving to, and taking away from, future level designers. Lix strongly encourages use of 16-wide blocks; ccexplore asks "couldn't we just violate the grid?" but (1) this would force ugliness in levels using brick or tile terrain that emphasises the separate blocks; (2) you then can't use the editor's snap-to-grid function, so placing the terrain becomes much more of a hassle, and there may be a lot of adjacent terrain that has to be moved slightly off-grid to accommodate.
And then there's my earlier point about consistency between the basher and digger. geoo has pointed out that usually, the basher is walking on ground that determines his starting height, so it doesn't matter how much leniency he has. Still, it would feel like a major inconsistency if the basher can get between steel blocks that are 16 pixels apart, and the digger can't do so at all.
So, for me, this is an absolute deal-breaker that prevents the 9/9 digger being a viable option. Others may feel less strongly about this issue.
This is also a strike against the 7/9 digger, which allows digging down 16-wide terrain between steel, but only with a frame-perfect starting point. New players may miss a few times and give up, not realising there is a viable starting point.
Razor's Edge digger: Simon brought this up in chat. With a 5/9 or lower digger, you can dig down along the edge of 1-wide steel, and remove terrain on the other side. geoo's "construction" set does actually include 1-wide steel, so we can't write this off as "academic, will never meet in practice".
The miner and bomber can also remove terrain on the other side of narrow steel, so it's not necessarily a big problem. However, it feels especially weird with the digger, because you can continue down the edge of 1-wide steel and remove terrain on the other side all the way. Sooner or later, someone will (re)discover this and use it in an intended solution.
Cuber plugs digger hole: This is geoo's objection. With a 7/9 or lower digger, you can dig close in to steel so the steel "covers" part of the dig pit, allowing the remainder to be plugged with a cuber. This is especially bad for the 7/9 digger, because only a single starting pixel allows this trick, making it harder to discover (and, again, sooner or later someone will use it in an intended solution).
I don't think allowing this trick is necessarily bad, but I do agree with the point that it's bad to allow tricks that are inherently precise. (Then again, one is used to cuber placement being precise, for instance when trying to cover a gap with the minimum number of cubers. Levels like Cubic Interpolation give enough leeway that this isn't required, but it's always encouraged, because each cuber kills one lix.)
"Changing of the Guards" scenario: You have to dig close to steel, and some other factor encourages getting as close as possible, even if the level actually gives some leeway so that it's still possible to solve if you start your digger a bit further out.
This seems to be much more common in actual practice than the "steel on both sides" scenario, and it's an argument against the 7/9 and 9/9 diggers. With more leniency, you can start just after the walker lix has crossed from steel onto normal terrain. With less leniency, you have to learn the amount the lix has to walk before she can start digging. This feels like fiddliness we don't want (and note that it has no equivalent for the basher).
In summary, my vote is for the 5/9 digger, but in that case we'd have to accept the "Razor's Edge" problem. I wouldn't mind going with the 7/9 digger if others strongly want to avoid that problem, but I strongly object to 9/9.