Code Review

I've analyzed this PR and the changes look good overall. The fix correctly addresses the issue where the stone removal clock wasn't resetting properly by using the more specific stone_removal_expiration field when available.

Minor observation:

The stone_removal_mode field added to AdHocClock interface doesn't appear to be used anywhere in the codebase. While this might be intentional (for future use or documentation purposes), you may want to verify if it's needed as part of this change.

The logic using ?? operator is correct and maintains backwards compatibility when stone_removal_expiration is not set.

My Claude added the other field because technically this is what the server uses to indicate that the stone_removal_expiration is valid. Goban infers this indirectly instead as it happens. But the server is sending it, so it is in the JSON that's being described.

Awesome, thanks for digging into that and getting it fixed!