I have to operate into these sorts of runways so on reflection I think this is how I do it...
There seems to be a some competing requirements in the situation:
1. Some reference for pitch attitude control ie the end of the runway or some point along it far enough away to remove the perception of rate of longitudinal closure
2. Some reference to determine rate of vertical closure ie sink (or the opposite....
)
3. Some reference that might allow you to anticipate #2.
I think that with a flat (note: not necessarily level) runway then looking some distance down the runway can accommodate all those tasks. As soon as the runway becomes undulating, crested or dipped then that single reference point becomes inadequate.
1. Pitch reference: I use some length down the runway OR the end of the runway if it's short OR the visible horizon if it's crested.
This gives me an attitude reference point that is relatively constant. If crested then as I lose the crest I transfer to using further along the runway.
2. Sink rate: I tend to look much closer. Not very far in front of the cowling (10m? 20m? 50m? Not sure, but it's quite close). I also pay more attention to my peripheral vision.
Using the perceived sink rate I modify my 'normal' flare & hold off, pitching a bit faster if I'm sinking fast, less so if I'm not sinking enough. The goal is to approx. fly parallel to the surface regardless of its slope then achieve a controlled sink.
This is a very short term 'reaction' type input.
3. Anticipation: If I can see the runway is falling away/going to fall away then I plan not to hold off as much, allowing the a/c to descend. Similarly, if I can see the runway rising ahead of me then I pitch more with the aim of flying parallel to the surface (not the horizontal).
This is a slightly longer term modifier to #2 above.
The end result is that I scan ahead & back throughout the process.
Additionally I modify my power usage. If a downslope is the first 'problem' then I'll reduce power earlier &/or faster. If an upslope I'll reduce power later &/or more slowly. If quite a steep up slope initially I'll sometimes fly a quite shallow approach gradient (obstacles permitting of course) to reduce the amount of flight path change needed.
Another 1 or 2 kts above 1.3 Vs can be useful to give a **little** extra energy for the extra pitch up for the upslope. I don't mean 10 kts or whatever, but just the aiming for the fast side of the required speed instead of the slow side.