Rrd update crash - ifHCOutBroadcastPkts outsized

When I poll some ports on my device, the value return is too big for the rrd to handle, and rrdcached crashes.
I tried deleting the rrd & restarting rrdcached, but observium recreates it with the same max value, and the error comes back.

When rrdtool crashes, all graphs are down, which is pretty annoying.

Debug :

./poller.php -h 3 -m ports -d
[...]
ifXEntry DEBUG: SNMP Auth options = -v2c -c 'pouetpouet'

CMD[/usr/bin/snmpbulkwalk -v2c -c 'pouetpouet' -Pu -OQUs -m IF-MIB -M /opt/observium/mibs/rfc:/opt/observium/mibs/net-snmp 'udp':'mydevice.domain.tld':'161' ifXEntry]

EXITCODE[0]
CMD RUNTIME[0.0177s]
STDOUT[
ifName.1 = vmnic0
ifName.2 = vmnic1
[...]
ifHCOutMulticastPkts.6 = 0
ifHCOutMulticastPkts.7 = 0
ifHCOutBroadcastPkts.1 = 101115
ifHCOutBroadcastPkts.2 = 18446742978492891134
ifHCOutBroadcastPkts.3 = 63455
ifHCOutBroadcastPkts.4 = 538
ifHCOutBroadcastPkts.5 = 0

// The problematic value is ifHCOutBroadcastPkts.2

[...]
RRD CMD[update /opt/observium/rrd/mydevice.domain.tld/port-2.rrd N:2222411483:112999116:0:0:75870324490:73554100427:373958:1.8446742978497E+19:0:0:0:173621:18446742978492891134:200337:4540969 --daemon unix:/var/run/rrdcached.sock]
RRD RUNTIME[0.0011s]
RRD STDOUT[OK u:0.00 s:0.00 r:0.41]
RRD_STATUS[TRUE]

// This command generate this error in syslog :
Oct 27 09:11:30 SUP11 rrdcached[12750]: queue_thread_main: rrd_update_r (/opt/observium/rrd/mydevice.domain.tld/port-2.rrd) failed with status -1. (/opt/observium/rrd/mydevice.domain.tld/port-2.rrd: not a simple signed integer: '1.8446742978497E+19')

1. rrdtool info /opt/observium/rrd/mydevice.domain.tld/port-2.rrd|grep -i OutBroadcastPkts
   ds[OUTBROADCASTPKTS].index = 12
   ds[OUTBROADCASTPKTS].type = "DERIVE"
   ds[OUTBROADCASTPKTS].minimal_heartbeat = 600
   ds[OUTBROADCASTPKTS].min = 0.0000000000e+00
   ds[OUTBROADCASTPKTS].max = 1.2500000000e+10
   ds[OUTBROADCASTPKTS].last_ds = "18446742978492891134"
   ds[OUTBROADCASTPKTS].value = NaN
   ds[OUTBROADCASTPKTS].unknown_sec = 68