mdadm not kicking faulty disk from raid, why?

[Harkonnen]

So, today one of the disk in my raid-1 array started failing, but mdadm didnt kick it from raid, continued to try to read from that disk, until eventually some error didnt slip to the upper level and caused root filesystem to turn ro.

this is how it looks in logs, without actual ATA errors lines

Aug 14 17:17:05 nas kernel: [99299.754478] print_req_error: I/O error, dev sdd, sector 80716384
Aug 14 17:17:05 nas kernel: [99299.754542] md/raid1:md0: sdd3: rescheduling sector 79692384
Aug 14 17:17:09 nas kernel: [99303.855795] print_req_error: I/O error, dev sdd, sector 80716384
Aug 14 17:17:10 nas kernel: [99303.999038] md/raid1:md0: read error corrected (8 sectors at 79692384 on sdd3)
Aug 14 17:17:10 nas kernel: [99303.999053] md/raid1:md0: redirecting sector 79692384 to other mirror: sdd3

and then it continues until next error where similar pattern is repeated. I expect that mdadm would kick the faulty mirror from raid after errors...

[LE_746F6D617A7A69]

You forgot to tell what raid configuration (/proc/mdstat and mdadm -D <array>) and what FS You're using.
Also, It would be nice to see a full SMART report for the sdd drive.
What is that "nas kernel"?

But besides that: I can't see anything strange in that log -> raid arrays at level 1 and higher are supposed to recover damaged sectors from backup. I think that the problem is somewhere else...