How to encode multiple voices on a multistaff part

[rpatters1]

The reference document states:

Within a given measure, no two sequences may share the same value for voice.

However, how does this apply in multi-staff instrument situations? Possibilities seem to be

• don't use voices for multi-staff instruments
• the voice label uniqueness requirement should be per staff
• the staff number should be included in the voice label (barf)

Any advice is welcome.

[rpatters1]

It would be nice to get an answer on this. I am particularly interested to know if the uniqueness requirement for voice labels in sequences is per-staff. That seems like the most logical to me, but if so it needs to be clarified.