Bump Marten to 9.7.3 (+ JasperFx 2.9.9) — fix tenant-partitioned composite catch-up under managed distribution

[erdtsieck]

What

Bump the pinned Marten 9.6.0 → 9.7.3 and JasperFx 2.8.2 → 2.9.9 (the version 9.7.3 depends on), and add a regression test for composite projection catch-up under UseWolverineManagedEventSubscriptionDistribution + Marten UseTenantPartitionedEvents.

Why

On the pinned Marten 9.6.0, a composite projection never catches up under the Wolverine-hosted daemon with tenant-partitioned events — it stays stuck at a premature 0 → 1 optimized rebuild, while a plain SingleStreamProjection in the same store catches up and Marten's own daemon handles the composite fine.

Root cause is not in Wolverine's distribution layer. Marten 9.6.0 predates marten#4712 (high-water Timestamp under tenant-partitioned events): the store-global HighWaterAgent reads a default(DateTimeOffset) = 0001-01-01 timestamp, computes a bogus safe-harbor (0001-01-01 + StaleSequenceThreshold), and never advances:

Starting single-pass composite rebuild for CsComposite:All over events 0 to 1
Shard 'CsPlain:All': Executed updates for Event range ... 0 to 30
Finished ... CsComposite:All, proceeding to normal, continuous operation
High Water agent is stale at 30 ... skipping gap to events marked after 01/01/0001 00:00:03
Daemon projection high water detection was not able to determine a safe harbor sequence, will try again soon   (forever)

Clean A/B, only the Marten version changed:

Marten	Result
9.6.0 (current pin)	hang ~90s, composite stuck at 1, stale … 01/01/0001
9.7.3	PASS in ~4s

The full wolverine.slnx builds clean against the bump (0 warnings / 0 errors).

Closes #3084. Supersedes #3083 (which assumed a Wolverine-layer bug).

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

Adds a regression test to ensure composite projections catch up correctly under Wolverine-managed event subscription distribution when tenant-partitioned events are enabled, and updates Marten/JasperFx dependencies to versions containing the underlying high-water mark fix.

Changes:

• Added a new regression test covering composite projection catch-up across multiple managed tenants with UseTenantPartitionedEvents.
• Updated JasperFx packages to 2.9.9.
• Updated Marten packages to 9.7.3.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
src/Persistence/MartenTests/Distribution/composite_catches_up_under_managed_distribution_with_partitioning.cs	Adds a regression test validating composite projections catch up for tenant-partitioned events under Wolverine-managed distribution.
Directory.Packages.props	Bumps JasperFx and Marten package versions to pick up fixes needed by the regression scenario.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[erdtsieck]

Note on the failing CI checks — unrelated to this bump

The red build and Run Marten Tests checks fail at restore, not in any test:

error NU1903: Warning As Error: Package 'MessagePack' 3.1.3 has a known high severity
vulnerability, https://github.com/advisories/GHSA-hv8m-jj95-wg3x

This is a newly-published advisory for MessagePack 3.1.3, which is pinned repo-wide in Directory.Packages.props and is not touched by this PR. With TreatWarningsAsErrors=true the NuGet audit warning becomes a build error. main's CI was green on 2026-06-11 and would now hit the same NU1903 if re-run, since the audit evaluates live advisory data at restore time — only the jobs that pull in Wolverine.MessagePack fail; every job that doesn't passes.

So the failures are orthogonal to the Marten 9.6.0 → 9.7.3 bump. Locally the full wolverine.slnx builds clean and the new test passes in ~4s with -p:NuGetAudit=false. Resolving the advisory (e.g. MessagePack 3.1.3 → 3.1.7) is a separate, repo-wide change and is intentionally left out of this PR.

[jeremydmiller]

Rebased onto current main and merged via #3105 (your authorship preserved via cherry-pick). main was already at JasperFx 2.9.10, so only the Marten 9.6.0 → 9.7.3 bump and your regression test were carried over. Thank you @erdtsieck!