Previous patch solved the problem with a half measure: it simply
copied the missing part of the reconnection flow from the connection
flow (namely, saving the DBI handle). It's better to just use the
same code.
Short term benefit: the last_connection_check timestamp
will be set at connection time, instead of at the next call after
reconnect, which allows to spare a connection check on the next
call.
Long term benefit: fewer chances of diverging further.
Previous patch solved the problem with a half measure: it simply copied the missing part of the reconnection flow from the connection flow (namely, saving the DBI handle). It's better to just use the same code.
Short term benefit: the last_connection_check timestamp will be set at connection time, instead of at the next call after reconnect, which allows to spare a connection check on the next call. Long term benefit: fewer chances of diverging further.