concurrent datalad operations?

[mgxd]

When concurrently converting across multiple participants while using --datalad, heudiconv terminates before completion due to git index.lock files and the dataset is not put under version control.

Perhaps OpenNeuroOrg/datalad-service can be some inspiration

[yarikoptic]

Yeap, a known limitation of the git and even you greater degree of datalad. No easy way around unfortunately ATM.
OpenNeuro datalad backend unlikely to have this issue since unlikely had multiple datalad invocations done on parallel in the same dataset, and they don't manage a hierarchy of the datasets.
They only workaround is to not use --datalad mode ATM and perform datalad add later for the entire result. This would though cause some files lacking their metadata tags

[satra]

one possibility may be to stage the info into separate uuid branches and then merge it later in one go after conversion.