Zabbix creates "down"/"lost" events even if the host is available.
Sometimes (I didn't find a pattern yet) down hosts will become "lost" leading to a discovery event the next time they show up (instead of a normal "up").

Maybe it's just a misunderstanding, but I couldn't find any documentation when being "down" will lead to be "lost".

// SQL query to get the raw data
SELECT e.value, FROM_UNIXTIME( e.clock, '%Y-%m-%d %H-%i-%s' ) AS date
FROM `events` AS e
LEFT JOIN dservices AS ds ON e.objectid = ds.dhostid
WHERE e.source =1
AND e.object =1
AND ds.ip LIKE '10.222.111.58'
ORDER BY ds.ip DESC , e.clock DESC

// laptop
0 2012-04-18 09-18-16
2 2012-04-18 09-18-16
1 2012-04-18 09-07-34
1 2012-04-18 08-56-53
3 2012-04-18 08-56-53
0 2012-04-18 08-46-11
// more UPs
2 2012-04-18 08-03-25
0 2012-04-18 08-03-25
1 2012-04-18 07-52-43
// more DOWNs
1 2012-04-17 16-52-12
3 2012-04-17 16-52-12
// more UPs
0 2012-04-17 16-09-24
2 2012-04-17 16-09-24
1 2012-04-17 15-58-42
// more DOWNs
1 2012-04-16 16-56-14
3 2012-04-16 16-56-14
2 2012-04-16 16-45-33
0 2012-04-16 16-45-33
1 2012-04-16 16-34-51
1 2012-04-16 16-24-09
3 2012-04-16 16-24-09
0 2012-04-16 16-13-28
0 2012-04-16 16-02-46
0 2012-04-16 15-52-04
2 2012-04-16 15-41-23
0 2012-04-16 15-41-23
1 2012-04-16 15-30-40
1 2012-04-16 15-19-58
1 2012-04-16 15-09-16
3 2012-04-16 15-09-16
0 2012-04-16 14-58-34
// more UPs
0 2012-04-16 12-28-50
2 2012-04-16 12-28-50