Odd Paste Special->Reduce behaviour

Hello, I have a 4 part choir setting, notated on 4 staves. I am doing a reduction of this music to a two stave organ.
All 4 singers use up-stem voice 1.
Now I select Soprano and Alto staves, copy and then Paste Special->Reduce onto the upper organ staff.
I get the music nicely distributed, the Soprano upstem and the Alto downstem.
Very nice, the same procedure with Tenor and Bass.
This was in the morning.
As the day progressed, suddenly Soprano and Alto ended up in one voice (upstem voice 1). Ok, they went parallel, so it still looked ok.
But even later in the day Soprano and Alto (and Tenor and Bass) both ended up in two different downstem voices, all stems pointing down. This just by using the Paste Special->Reduce function.
I could not get back to the results from this morning. Even after relaunching Dorico…
I went the longer route by copying/pasting all 4 voices separately.
The results are fine but it takes double the time. Am I overlooking something?

[follow up]
Today I resumed my arrangement.
Reducing Soprano and Alto into the top organ staff did work again.
I don’t know why, but it just gave good results again.
After 30 minutes the behaviour changed. I did inspect, what was wrong - the Reduce function produced the Alto to end up in upstem voice one and the Soprano in upstem voice two, why?
I then used the function (after selecting) Swap Voice content. Then filtered the Alto and changed the voice to downstem 1.
I also tried another approach to save time: I selected all 4 singer’s parts, copied and paste special->reduce onto two organ staves. Sometimes good results, sometimes not. I used the Swap Voice content a few times.
Sometimes it fixed the stem directions, sometimes not. Even a few times the Swap Voice Content let both voices completely disappear . This was really strange. If I swapped in smaller batches, it would eventually work.
It feels as if there is some incomprehensible AI running in the background…

This is very curious, but in order to say anything about it we’ll need to see concrete reproducible examples. Anecdotal evidence doesn’t really do anything.

1 Like

ok, next time I’ll do a screen recording, promised.