chore(deps): tier 4 cascade — consume tiers 0-3, StackExchange.Redis 3.1.13

[poxet]

Tier 4 of the cascade update, and the last Toolkit repo in scope. Consumes everything below it.

Internal — consuming tiers 0-3

Package	From	To
Tharga.Console	4.2.0	4.2.1 (Sample)
Tharga.Blazor	2.2.3	2.3.0 — brings Radzen.Blazor 11.x, a major
Tharga.Mcp	1.0.0	1.1.0 — brings MCP SDK 2.2.0 (stateless HTTP transport by default)
Tharga.MongoDB	2.14.1	2.14.3 — brings MongoDB.Driver 3.11.0

All verified by an actual dotnet restore, not a flatcontainer listing. Tharga.MongoDB 2.14.3 needed a wait: flatcontainer listed it while restore still resolved 2.14.2 as newest.

Note the two behaviour-bearing ones. Tharga.Cache.Blazor picks up the Radzen major transitively, and Tharga.Cache.Mcp picks up the SDK's stateless-transport default. Both are why those packages shipped as minors rather than patches.

External

Library

• StackExchange.Redis 3.0.7 → 3.1.13 (Tharga.Cache.Redis) — six patch levels
• Microsoft.Extensions.Hosting.Abstractions 10.0.9 → 10.0.11
• Microsoft.SourceLink.GitHub 10.0.300 → 10.0.400 (five projects; PrivateAssets=all, build-only)

Polly 8.7.0 is current and unchanged.

Test tooling (does not ship, all four test projects)

• Microsoft.AspNetCore.Mvc.Testing 10.0.9 → 10.0.11
• Microsoft.NET.Test.Sdk 18.7.0 → 18.9.0

Verification — and a flaky suite, measured rather than assumed

dotnet build -c Release --no-incremental — 1 warning, 0 errors. The warning is CS0162 (unreachable code) in Tharga.Cache.File/File.cs:63, confirmed pre-existing by rebuilding that project without this change.

dotnet test -c Release:

• Tharga.Cache.Tests — 477/478, with 1-2 failures in FetchDataThrottleTests
• Tharga.Cache.Redis.Tests — 5/5 passing
• Tharga.Cache.File.Tests — 2 skipped

The throttle failures are pre-existing flakiness, and I measured that on clean master rather than assuming it. Method: stashed every dependency change, rebuilt, and ran the filtered class four times on unmodified master. Result: pass, fail, fail, fail — it fails most runs with no changes at all.

The assertions are exact event counts under parallel load, so they lose races:

Expected dataGetEventCount to be greater than or equal to 8, but found 7.
Expected monitorGetEventCount to be 10, but found 9 (difference of -1).

Filed as a new Important/Medium backlog item. Not fixed here — this is a dependency PR — but worth flagging plainly: a suite that fails most runs cannot answer "did my change break something?", and establishing that these bumps were innocent cost a four-run baseline.

Left filed, deliberately

Both open issues are untouched, since this pass was scoped to dependencies:

• #56 — FluentAssertions ships as a public dependency. Confirmed genuine while here: FluentAssertions 8.10.0 is referenced from Tharga.Cache/Tharga.Cache.csproj, the shipping library, not a test project — so its paid Xceed licence does propagate to consumers. Worth prioritising.
• #55 — AddCache is not safe to call concurrently (static registration dictionary races).

No source changes, no public API change, no MAJOR_MINOR bump (stays 1.0).

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[github-actions]

Released as v1.0.2 — https://github.com/Tharga/Cache/releases/tag/1.0.2