When a voice splits in two voices, and then go back together, I usually
use
<<
{ \voiceOne ... }
\new Voice { \voiceTwo ... }
>>
Although this makes a good output (check block 3 in the attached image),
it makes quarter pauses to go up after this block.
I tried some combinations of stemUp/stemDown, voiceOne/VoiceTwo but I
can't get a good one. One the voiceOne/voiceTwo render correctly, but it
is the only that messes with the placement of pauses after it.
I can tweak some of the previous blocks (1 or 2) forcing stems and slurs
to go up or down, but I do not think that's the better approach.
So, I am missing something.