Provider destined to fail activity creation due to corrupted downloaded image

[grisha87]

Steps to reproduce

1. Start deploying an activity on a provider that doesn't have a specific image in the cache
2. Terminate the activity during the download before it completes
3. Try to run the activity again on the provider with the same image

Actual result

The provider fails to start the activity reporting that the image is corrupted. Error: Preparing activity failed. Error: Local service error: Transfer error: Downloaded VM image is corrupted: calculated hash 869e1cf53a3a707b1731b48c820f2c637d4754c4365d83d50cf03e7b differs from the expected one 49dacaf80fdf1aadf70afa1cfbc46d3b6c694382e1a53eaaec9afcd1

Expected result

• I should be able to run the activity on the provider
• The broken image once identified could be deleted and re-downloaded
• If the download is interrupted, the image should be also removed from the cache

Context

In my Golem Node Validator dataset I was able to identify 10 providers suffering from this error.

[grisha87]

It appears that this might be fixed in 0.16 with the changes introduced by @nieznanysprawiciel .