Spurious sounds from delay?

Sometimes delay surprises me. E.g., this simple code

d1 $ s "sn:1 "
      # delay 1
      # delayt (16/9) # lock 0

sometimes produces the expected phasing (the delayed sound is exactly one cycle after the original) but sometimes, something extra (at half a beat) https://gitlab.imn.htwk-leipzig.de/waldmann/computer-mu/-/blob/master/tidal/code/spurious-delay.ogg

While trying to answer "what is the time unit for delayt", I learned about lock:

The behaviour of the code is the same with # delayt 1 # lock 1 (equivalent for default cps) in the last line.

Spurious delays only seem to occur for events on cycle boundaries. Sometimes I think that some delays are missing.