CTM-292 Discontinue fruitless resolution of DRS files in localization_optional

[aednichols]

Description

Before

Users can supply a mix of DRS and GCS files in the localization_optional section. We attempt to resolve the DRS files to GCS paths.

Resolve success: skip localization and provide the gs:// path to the task

Resolve failure: behave as if the file was not localization_optional and download the file with the DRS localizer

What changed

Some years ago, DRS discontinued resolving to GCS paths. It now always provides signed URLs.

Therefore, attempting to resolve DRS paths to GCS is dead code.

We're left with a weird mechanism that resolves DRS files in localization_optional but always localizes them.

After

If a user provides a DRS path in localization_optional, we will not attempt GCS resolution.

We will simply localize all DRS files.

Possible breakage

Workflows will break if there are still DRS files out there that resolve to GCS paths, and the user relies on Cromwell supplying a gs:// path to the task. We don't think there are such files in use, but it is not possible to verify.

localization_optional does not mean "never localize". OpenWDL and Cromwell document that WDLs need to be written to work whether the engine localizes or not.

The workaround is to follow the spec and write tasks that also work with local paths; or switch away from DRS to GCS if tasks absolutely require gs:// access.

Why do we care about this now

The process of trying to DRS-resolve an extreme amount (7M) of localization_optional files caused Cromwell outages the week of 2025-12-08. It turns out we shouldn't be doing that anyway.

Release Notes Confirmation

CHANGELOG.md

• I updated CHANGELOG.md in this PR
• I assert that this change shouldn't be included in CHANGELOG.md because it doesn't impact community users

Terra Release Notes

• I added a suggested release notes entry in this Jira ticket
• I assert that this change doesn't need Jira release notes because it doesn't impact Terra users

[aednichols]

Do not merge because: ready for code review but still need to double-check requirements with Michael Baumann

[LizBaldo]

Thanks, the changes make sense to me, but better safe than sorry, good call on double checking with Michael.

[aednichols]

@mbaumann-broad / @aednichols sync:

Neither of us had heard of this niche use case. We don't think it sees any significant usage.

If users want localization optional with DRS, we should pave a path forward for them by adding L.O. support for signed URLs. This is preferable to going backwards and investing in the deprecated DRS-to-GCS pathway.

@aednichols to check metrics to see if we have any data on usage of this feature.

@mbaumann-broad to check with Alex B. to see if he knows of any users in his network.