+<a-. c\prall e>1\sfz c'4 q2 r8 q8 |
+q2 c, |
+@end lilypond
+
+To have some of them retained, the @code{\chordRepeats} function can be
+be called explicitly with an extra argument specifying a list of
+@var{event types} to keep unless events of that type are already
+present on the @code{q} chord itself.
+
+@lilypond[verbatim,quote]
+\relative c'' {
+ \chordRepeats #'(articulation-event)
+ { <a-. c\prall e>1\sfz c'4 q2 r8 q8-. } |
+ q2 c, |
+}
+@end lilypond
+
+Here using @code{\chordRepeats} inside of a @code{\relative} construction
+produces unexpected results: once chord events have been expanded, they
+are indistinguishable from having been entered as regular chords, making
+@code{\relative} assign an octave based on their current context.
+
+Since nested instances of @code{\relative} don't affect one another,
+another @code{\relative} inside of @code{\chordRepeats} can be used for
+establishing the octave relations before expanding the repeat chords.
+In that case, the whole content of the inner @code{\relative} does not
+affect the outer one; hence the different octave entry of the final note
+in this example.
+
+@c Without \new Voice, implicit voice creation does the dumbest thing.
+@lilypond[verbatim,quote]
+\new Voice
+\relative c'' {
+ \chordRepeats #'(articulation-event)
+ \relative c''
+ { <a-. c\prall e>1\sfz c'4 q2 r8 q8-. } |
+ q2 c |
+}