Help: Answers are getting cut off mid-sentence. Is there a token limit?

I've been putting NotebookLM through its paces for the last three weeks, primarily using it to analyze and structure complex campaign performance reports. I've hit a consistent and frankly, maddening, roadblock: the assistant's responses are being truncated, often at what seems like a completely arbitrary point mid-sentence. It's not just an occasional hiccup; it's a systematic failure that renders longer analytical summaries or comparative breakdowns useless.

The context is crucial here. I'm not asking for sonnets. I'm feeding it a dense source document—say, a multi-channel attribution report export—and asking a compound question like, "Based on the data in Source 1, compare the assisted conversion paths for Campaign A and Campaign B, then hypothesize why the lead scoring threshold might be misaligned for Campaign B's MQLs." The response will start strong, laying out a clear comparison, begin the hypothesis... and then just stop. No period. No completion of the thought. It's like the connection drops, but everything else is fine.

This screams of a hard token limit on output, not input. I've confirmed my prompts are well-formed and the source material is within the stated limits. The lack of transparency here is the real issue. If there's a 500-word or 1000-token cap on answers, fine, but I need to know so I can structure my prompts accordingly. Right now, it's a black box that fails silently at the worst possible moment—just as the insight is about to be delivered.

My immediate questions for the community are:
* Has anyone definitively mapped the cutoff point? Is it character-based, token-based, or something else?
* Are there known workarounds beyond the obvious "ask for shorter answers"? Sometimes the analysis *requires* depth.
* Most importantly, has Google acknowledged this as a known limitation, or is this potentially a bug specific to certain account types or document complexities?

Without understanding this constraint, the tool's utility for any serious marketing analysis—where conclusions matter—is severely compromised. You can't base a strategic pivot on half an answer.

--- M^2