When the "Single TTS Instance" option is on, and the system-wide default TTS voice is Microsoft David, with no profile or action-level overrides, as long as no other voices have been used, "David" will be used for "Say Something with Text-To-Speech" actions with the "Voice" option set to "Default".
However, as soon as a TTS action using a different voice (E.G. Microsoft Zira) has played, that voice will instead be used as the default TTS voice from that point onward (until VoiceAttack is restarted).
E.G.
Play a TTS action with default -> "David" plays
Play a TTS action with default -> "David" plays
Play a TTS action with default -> "David" plays
Play a TTS action with "Zira" -> "Zira" plays
Play a TTS action with default -> "Zira" plays
Play a TTS action with default -> "Zira" plays
Play a TTS action with "David" -> "David" plays
Play a TTS action with default -> "David" plays
Where it gets even weirder is if "Single TTS Instance is unchecked, as then, all voices previously used will, in order of appearance, become the default voice.
E.G.
Play a TTS action with default -> "David" plays
Play a TTS action with Ivona "Brian" -> Ivona "Brian" plays
Play a TTS action with Ivona "Russell" -> Ivona "Russell" plays
Play a TTS action with "Zira" -> "Zira" plays
Play a TTS action with default -> Ivona "Brian" plays
Play a TTS action with default -> Ivona "Russell" plays
Play a TTS action with default -> "Zira" plays
Play a TTS action with default -> Ivona "Brian" plays
Play a TTS action with default -> Ivona "Russell" plays
Play a TTS action with default -> "Zira" plays
This also occurs in v1.8.3, and at least as far back as v1.7.5
EDIT: Fixed in v1.8.3.18