Blocked behavior when server closes connection while downloading

[rngtng]

We've a problem with blocked operation when trying to downloading a non-existent file and therefor the serves closes connection. When I try same operation on shell, No 'file not found' msg is shown, instead connection closed right away.

Ruby code I execute.

file_name = "doesnotexists.file"
Net::SFTP.start(@host, @username, @options) do |sftp|
     data = sftp.download!(file_name)
end

Logs:

I, [2020-09-10T18:41:57.010051 #344]  INFO -- net.sftp.session[c1c]: sending open packet (0)
D, [2020-09-10T18:41:57.011263 #344] DEBUG -- socket[ba4]: queueing packet nr 9 type 94 len 44
D, [2020-09-10T18:41:57.012189 #344] DEBUG -- socket[ba4]: sent 112 bytes
D, [2020-09-10T18:41:57.070221 #344] DEBUG -- socket[ba4]: read 304 bytes
D, [2020-09-10T18:41:57.070756 #344] DEBUG -- socket[ba4]: received packet nr 9 type 96 len 28
I, [2020-09-10T18:41:57.071407 #344]  INFO -- net.ssh.connection.session[c08]: channel_eof: 0
D, [2020-09-10T18:41:57.073445 #344] DEBUG -- socket[ba4]: received packet nr 10 type 98 len 44
I, [2020-09-10T18:41:57.074479 #344]  INFO -- net.ssh.connection.session[c08]: channel_request: 0 exit-status false
D, [2020-09-10T18:41:57.075367 #344] DEBUG -- socket[ba4]: received packet nr 11 type 97 len 28
I, [2020-09-10T18:41:57.075652 #344]  INFO -- net.ssh.connection.session[c08]: channel_close: 0
D, [2020-09-10T18:41:57.076142 #344] DEBUG -- net.sftp.session[c1c]: sftp channel closed
D, [2020-09-10T18:41:57.077775 #344] DEBUG -- socket[ba4]: queueing packet nr 10 type 97 len 28
I, [2020-09-10T18:41:57.078294 #344]  INFO -- net.ssh.connection.session[c08]: host delete channel 0 which closed locally and remotely
D, [2020-09-10T18:41:57.079023 #344] DEBUG -- socket[ba4]: sent 96 bytes

Do we use/config the gem wrong? What's best way to solve this? We had idea first get directory listing, although for a large directory this takes a while. Other idea is to span an additional thread which kill the command after a while.

Any suggestions?

[rngtng]

Update: problem seems that server closes the connection, I see that @sftp.session.channels is empty, but the download event loop still active as on_close isn't called

[rngtng]

Update: the current workaround we found it to spawn a watch-thread which kills are timeout, see https://stackoverflow.com/questions/9894407/ruby-netsftp-transfer-interrupt-handling