In the end, I decided to suck it up and give To Say Nothing of the Dog the attention it deserved. Having made that decision, it took me only two days to polish off the second half (the first took me almost three weeks).
I was right about the difficulties I was having: they had nothing to do with the writing, and everything to do with the way I was reading it. This book is a really clever mystery, and I was missing that because I wasn't concentrating. I didn't get the big twist, but I got enough of it to feel like I hadn't been totally left behind. I'm quite certain all the clues were there to piece the whole thing together, if only I'd paid closer attention. Even better, the hints were so beautifully integrated with the story that they never once called themselves out.
A few times, the main character was just on the verge of working everything out. I'm not sure if it was purely a writerly trick, but every single time he reached that point, I felt like I was almost there too. Whether it was a trick or not, it was excellently done.
In fact, that's my overwhelming impression of the whole novel: it was really well orchestrated. The payoff had all the satisfaction of a good mystery solved, but it also had all the (guilty?) pleasure that comes with the resolution of a good period romance. It even managed to sustain the dry humour right to the very last line. To Say Nothing of the Dog is definitely going on my bookshelf, because I'm sure I'll read it again. Quite a turnaround from my thoughts just two days ago, hey?
I haven't spoken yet about the other two books by female authors that I've read this year -- Elizabeth Bear's Dust and Nancy Kress' Steal Across the Sky -- but I will say that this plan to read more SF written by women is really paying off.