Basically there is an issue with containers not being able to loading file symlinks within directory symlinks. And of course, this is how Nextflow passess objects between pipeline processes... While I've found some partial work arounds, I am now completely stuck when it gets to the part of #C062H82UCRG where I optionally generate

tar.gz

files. I'm getting increasingly down a horrible rabbit hole of hacks and workaround that I'm deeply unsatisfied with (very brittle/flaky etc), where I'm having to make a 'custom' tar module and a lot of nasty channel manipulation which basically isn't working very well. The original idea behind the functionality was to make it easier to upload the databases to remote locations that can be downloaded on the fly - but I'm coming to the realisation that I think that most people wouldn't actually do this and even if people are - these would normally be people stuff running in the cloud and thus would use S3 anyway, and in this case you wouldn't use a tar archive as the s3 protocol supports directories... So I'm now wondering if I should remove the tar.gz generation functioanltiy entirely.