New tiles not created when replag high

Currently, I've imported a data extract into a rendering database, and am updating it using openstreetmap-tiles-update-expire, doing an hour's worth every 10 minutes so as not to overload the server. Currently, the state of that according to the relevant log is:

[2016-01-25 12:17:01] 26404 start import from seq-nr 1761684, replag is 21 hour(s)

Using a slippy map if I browse to a previously unrendered area I see as expected

DEBUG: START TILE ajt 15 16080-16087 10696-10703, new metatile

and a metatile is created. If I do a "/status" call on the new data on the map I see:

Tile is clean. Last rendered at Mon Jan 25 12:19:05 2016. Last accessed at Mon Jan 25 12:19:05 2016. Stored in file:///var/lib/mod_tile/ajt/15/0/50/233/220/8.meta

The timestamp on the created metatile file is:

-rw-r--r-- 1 user user 292203 Jan 25 12:19 /var/lib/mod_tile/ajt/15/0/50/233/220/8.meta

I suspect that there's a problem in that that's "now" rather than "21 hours ago" (the age of the data according to replag). If there was a change to that data 6 hours ago, and I browse to that area on the map again, I'm not seeing tiles being re-rendered until an explicit /dirty call on the tile, even in a different browser. I see the effect fairly regularly because I call openstreetmap-tiles-update-expire between certain hours of the day only (for reasons we don't need to go into).

Would it be feasible to set the metatile timestamp to match the replag (and presumably the tile expiry information sent back to the browser too) , or have I got hold of completely the wrong end of the stick here?

EDIT: amended title to try not to be misleading.

Start with openstreetmap-tiles-update-expire and the /status and /dirty paths described in the report, then trace how replag, metatile timestamps, and browser expiry information are handled. Reproduce the high-replag scenario and determine the expected timestamp and re-rendering behavior before defining a change and its tests.