avoid race conditions in case of scaling

[hleb-rubanau]

we can check for presence of parent directory but absence of files in it

If it is the situation, entrypoint can wait for e.g. 1 minute and switch into ssl_enabled mode than... or crash if nothing had been changed

[hleb-rubanau]

Use case is not completely clear, as anyway having multiple containers writing to the same volume is a potential source of bugs (or simply impossible to set up) in most environments.

So, for now let's stick to 'single dispatcher' concept, avoiding replication until we need to