DISPATCH-1096 - priority sheaf cleanup bug

[mgoulish]

This is for the bug that Ganesh and Chuck saw while killing and restarting routers.
The test succeeded 10 times each (with 0 false results) in these 4 cases:
{ Debug, RelWithDebInfo } x { Should_Fail, Should_Succeed }

[codecov-io]

Codecov Report

Merging #384 into master will increase coverage by <.01%.
The diff coverage is 100%.

@@            Coverage Diff            @@
##           master    #384      +/-   ##
=========================================
+ Coverage    84.8%   84.8%   +<.01%     
=========================================
  Files          73      73              
  Lines       16462   16466       +4     
=========================================
+ Hits        13960   13964       +4     
  Misses       2502    2502

Impacted Files	Coverage Δ
src/router_core/router_core.c	92.18% <100%> (-0.23%)	⬇️
src/router_core/connections.c	93.17% <100%> (-0.11%)	⬇️
src/container.c	76.55% <0%> (-0.2%)	⬇️
src/router_core/core_test_hooks.c	88.06% <0%> (+1.7%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update fa03ea4...28f08a2. Read the comment docs.

[ChugR]

This patch clears up the too-many-links for sheaf error and improves my router network stability. FOr this I approve the patch.

But in tests before and after this patch my throughput went from 2600 msg/S to 200-300 msg/S. The performance drop needs to be addressed as a separate issue.

[ted-ross]

Exiting the process here is a bit heavy handed. I think it's sufficient to log the critical error.

[ted-ross]

This patch clears up the too-many-links for sheaf error and improves my router network stability. FOr this I approve the patch.

But in tests before and after this patch my throughput went from 2600 msg/S to 200-300 msg/S. The performance drop needs to be addressed as a separate issue.

I don't see any changes in this PR that should affect throughput performance.

[ted-ross]

This patch clears up the too-many-links for sheaf error and improves my router network stability. FOr this I approve the patch.
But in tests before and after this patch my throughput went from 2600 msg/S to 200-300 msg/S. The performance drop needs to be addressed as a separate issue.

I don't see any changes in this PR that should affect throughput performance.

I withdraw the above statement. There appears to be a memory leak that is causing performance (and other) problems.

[mgoulish]

I yanked that exit(1).
I thought I needed that to detect failure in my python test, but that is not true.
Tests all still did the right thing, in 10 trials each (40 total) in the set: {should_fail, should_succeed} * { DEBUG, RelWithDebInfo}