[On1x]

What and why

pm_lifecycle_suite starts every case at fc::time_point::now(), and the harness takes genesis from the same clock (init_genesis in database.cpp). So these 72 cases silently change the rules they run under on the day a date-gated hardfork activates.

HF16 (CHAIN_HARDFORK_16_TIME, 2026-10-31T00:00:00Z) requires a QC over the parent on every block, exempting only the activation block's child (block_needs_parent_qc, round-anchor clause). A chain whose first block is already past the fork date therefore dies at block 3 with No QC for head #2, before any case tests what it is about:

12:34:29  apply_hardfork ] HARDFORK 16 at block 13712
fatal error: pm_lifecycle_suite/leverage_profit_close_never_mints:
  cert.valid(): No QC for head #13713 yet: cannot hand over the turn

That failure was run 41054140379 (PR #211, which only touches share/vizd/vizd.sh). The long leverage cases were the first to cross the date inside one run — the same master commit was green at 08:21 UTC and red at 12:34 UTC. From 00:00 UTC tonight it is not one case and not two jobs: all 72 cases, in all three consensus_sim jobs.

Change

Pin the case clock two days before HF16 — after HF15, before HF16, i.e. exactly the rules that were live on 2026-10-30. Two days of headroom for the longest case (~11 h 26 min of virtual time); both bounds are compile-time constants, so the pair stays ordered forever. One helper plus a mechanical replacement of the 72 start lines.

This is the cure test_pm_audit_fixes.cpp already applied to its own time bomb (its header says so). test_agent_access.cpp pins the same way.

Why not "vote for head"

Voting, as the HF16-aware scenarios do, is not enough here: after the fork the slot shortens to CHAIN_BLOCK_INTERVAL_HF16 (1 s) and _generate_block asserts the producer is the validator scheduled for the slot of when. This file's block-count arithmetic (blocks = (grace + 240) / CHAIN_BLOCK_INTERVAL + 80 and friends) is written against 3 s blocks, so a run that legitimately crossed the fork would under-cover its deadlines. Pinning keeps the file honest: it tests PM lifecycle under pre-HF16 rules, and HF16 itself is covered by test_hf16_boundary.cpp / test_hf16_production_safety.cpp.

Scope check

Four scenario files take the real wall clock; the other three vote for head already or pin (see above), and the rest start at 2026-01-01 or epoch. This file was the only one that both starts at now() and never votes.

Verification

• single-TU -fsyntax-only (testnet defines, nice -n 19 ionice -c3): clean.
• CI on this PR is the positive control: the same hour that reproduced the failure must be green here. Note that before 00:00 UTC CI exercises the pre-fork path this fix pins to — the decisive check is the first post-fork run on master, where No QC for head must not appear.

Merge is the owner's, as usual.