Bugs when the secondary ring includes the *next* time period.

We currently attempt to include the next TP (in addition to the previous one) in the secondary ring. Most of the time we don't have a shared random value for the next TP (find_srv_for_time returns None when called with the start value of the next TP), so the next TP is excluded from the secondary ring (I think this is why #1155 (closed) didn't happen very often). However, sometimes find_params_for_time returns Some, and the secondary ring winds up containing both the previous and the next TPs.

The following are not clear to me:

• why does find_params_for_time ever return Some for the next TP? According to srv-spec, cosensuses only need to include the SRV for the current and previous TPs (so where do we get the SRV of the next TP from?)
• Should the secondary ring include the next TP at all?