Drill 13 — Scheduled session reset and session windows

[sylvioCampos]

Every config in this lab sets NonStopSession=Y with StartTime equal to EndTime, and config/exchange.cfg says why: "the session never rolls, so a drill is never interrupted by a scheduled reset."

That is correct for the drills and wrong for reality, and the repo admits it twice without ever showing it:

• docs/drills/06-seqnum-persistence.md: "B3 resets drop-copy sequence numbers on its own schedule, several times a week. ResetOnLogon=N is correct within a session's life, not across a scheduled reset — you have to know the venue's calendar."
• docs/drills/07-gap-and-resend.md repeats the same caveat.

So a reader finishes drill 06 believing ResetOnLogon=N is simply the right answer, and it is only right between resets. On the wrong morning that belief produces a client stuck demanding a resend of messages that no longer exist on a counter that has already gone back to 1.

What the drill should teach

• StartTime / EndTime / NonStopSession and what the engine actually does at a boundary
• The sequence numbers resetting at a scheduled rollover, and both sides agreeing about it
• The failure: reconnecting across a boundary with a stale expectation, and what that looks like on the wire — which is not the clean gap of drill 07
• Why ResetOnLogon is a per-connect decision informed by a calendar, not a value you set once

Implementation sketch

The drills need this to happen on demand rather than at 23:00, so: POST /admin/roll forcing the venue through a scheduled reset, plus a config profile using real StartTime/EndTime so the reader can see the settings that drive it.

Acceptance

• docs/drills/13-session-schedule.md
• A test covering a clean rollover and a client reconnecting with a stale expectation
• Drills 06 and 07 updated to link here instead of leaving the caveat dangling