Orchestration tests have unhandled vow rejections

[michaelfig]

Describe the bug

While working on #10686 (convert unhandled vow rejections to unhandled promise rejections), we discovered that a few orchestration tests started to fail with unhandled promise rejections.

To Reproduce

Steps to reproduce the behavior:

1. Search for the use of expectUnhandled( in test files of agoric-sdk/packages/orchestration.
2. You'll find it used in a few places:

$ cd agoric-sdk/packages/orchestration
$ rg 'expectUnhandled\('
test/exos/local-orchestration-account-kit.test.ts
118:test(expectUnhandled(1), 'transfer', async t => {

test/examples/staking-combinations.test.ts
35:test(expectUnhandled(1), 'start', async t => {

test/examples/send-anywhere.test.ts
260:test(expectUnhandled(1), 'failed ibc transfer returns give', async t => {
$

3. If you run those tests individually, unhandled promise rejections are reported by AVA.
4. The use of expectUnhandled allows the tests to return success, even though AVA would normally fail.

I would expect these calls to the expectUnhandled(N) macros to be unnecessary.

Platform Environment

• what OS are you using? what version of Node.js? Agoric CI
• is there anything special/unusual about your platform? Yes, it's CI
• what version of the Agoric-SDK are you using? (run git describe --tags --always) a version that has fix(vow): report unhandled vow rejections on collection or upgrade #10686 merged in it

[michaelfig]

It looks like something in the mocked transfer timeout pathway (a rejected vow instead of a fulfilment ack) is not being handled by the orchestration package.

[0xpatrickdev]

@michaelfig thanks for the pairing today.

The 3 unhandled rejections are isolated to three places where we are incorrectly simulating a MsgTransfer timeout for a local-orchestration-account. We should have been simulating the timeout using the vtransfer bridge.

This commit fixes one of the tests, and leads us (+ @michaelfig ) to believe a MsgTransfer timeout will properly propagate as a Vow rejection for localOrchAccount.transfer(). As such, there's no impact for production.

I've relabelled this issue as a P2. The above commit needs some refactoring (alluded to in the commit message) before it can land on master.