I cannot test this using the original reproducer because the initscript seems to exit as soon as it finds that another process is hanging on the socket ... any hints?

now this is interesting - after modifying the reproducer to initialise the database at first, i.e. starting and stopping the daemon before, the daemon just starts, no matter that /var/lib/mysql/mysql.sock is used by nc ...

to be precise:

:: [ 12:25:38 ] :: [   PASS   ] :: File /var/lib/mysql/mysql.sock should exist 
srwxr-xr-x. mysql mysql unconfined_u:object_r:mysqld_var_run_t:s0 /var/lib/mysql/mysql.sock
:: [ 12:25:38 ] :: [  BEGIN   ] :: Socket opened by nc should contain /var/lib/mysql/mysql.sock :: actually running 'lsof -U -a -c nc | grep '/var/lib/mysql/mysql.sock''
nc      4873 root    3u  unix 0xffff99f38b7b6c00      0t0 63649 /var/lib/mysql/mysql.sock

...

:: [ 12:25:43 ] :: [   FAIL   ] :: Trying to start mariadb (should fail) (Expected 1, got 0)

...

:: [ 12:25:53 ] :: [   FAIL   ] :: Checking mariadb status (should not be running) (Expected 3, got 0)
:: [ 12:25:53 ] :: [   PASS   ] :: File /var/lib/mysql/mysql.sock should exist 
srwxrwxrwx. mysql mysql system_u:object_r:mysqld_var_run_t:s0 /var/lib/mysql/mysql.sock
COMMAND  PID  USER   FD   TYPE             DEVICE SIZE/OFF  NODE NAME
nc      4873  root    3u  unix 0xffff99f38b7b6c00      0t0 63649 /var/lib/mysql/mysql.sock
mysqld  5168 mysql   15u  unix 0xffff99f38ae99800      0t0 64204 /var/lib/mysql/mysql.sock


so, obviously, mariadb overlays the socket file somehow ...

in /usr/libexec/mariadb-wait-ready there is now connect-timeout added:

RESPONSE=`/usr/bin/mysqladmin --no-defaults --connect-timeout=2 --socket="$socketfile" --user=UNKNOWN_MYSQL_USER ping 2>&1`

re comment 7, 8 - there are a few other issues, e.g. bug 1598095

Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory, and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

https://access.redhat.com/errata/RHSA-2018:2439