Multiple thought data blocks are merged together

It is quite common in an agent's execution logic to have multiple segments of thought within a single conversation.

When my agent performs behaviors similar to this sequence:
```
thought -> message -> tool_call -> thought -> review (which is actually a type of thought)
```
all the thoughts get concatenated together.

Upon checking, it seems that the ACP protocol itself does not yet support messageId, which is the root cause of this issue.

Could you please advise if there are any alternative solutions that can present multiple segments of thought and multiple tool_calls in a single conversation—without concatenation, and in sequential order?

Thank you very much indeed.

No file or test is named. Start by tracing how the VS Code extension handles ACP thought, message, tool_call, and review segments, then examine the protocol's messageId limitation. Done means distinct thought segments and tool calls remain sequential rather than being concatenated, or the limitation and available workaround are documented.

Written by the indexing model from the issue text.