MacBookAir kernel_task at 300% and rising [duplicate]

I have a MacBookAir5,1 connected to a Thunderbolt Display.

When it gets very hot, PID 0 "kernel_task" uses ever more CPU until reaching 340% and slowing the Mac down sufficiently to make me restart it.

I found a proposed solution at this site but it says to "move the ‘plist’ file for the model identifier we discovered earlier out of that directory" which does not apply to me because my installation of Mountain Lion does not have a plist for my 2012 MacBook Air. system_profiler tells me that I have a MacBookAir5,1.

I tried a few things. I figured that since overheating appears to cause this, I should try not to overheat the poor Mac. However, I neither understand why Mac OS X' reaction to overheating is to use 340% of CPU time nor did I manager to keep the lid open and still convince the Airbook to use the Thunderbolt Display as its only display.

I tried running from battery, assuming this would make the CPU run slower and not overheat, but when I remove the power cable from the Airbook, the Thunderbolt Display just turns black, only recovering after several minutes when I reconnect the power cable.

What's going on?

Could you have a look at the 2 main system logs to see what your MacOS may be busy at. With a terminal or an xterm window, the 2 following commands fired when the problem appears may help you diagnose it:

tail -f /var/log/kernel.log

tail -f /var/log/system.log

top -o cpu

and grab a copy of the topmost 10 lines to put them back within your original question as here:

Processes: 50 total, 3 running, 47 sleeping, 243 threads                                                                                                                                       12:18:07
Load Avg: 1.40, 1.53, 1.28  CPU usage: 5.97% user, 7.14% sys, 86.88% idle    SharedLibs: 5044K resident, 9292K data, 0B linkedit. MemRegions: 10733 total, 769M resident, 15M private, 1479M shared.
PhysMem: 755M wired, 1363M active, 641M inactive, 2760M used, 1207M free. VM: 120G vsize, 1041M framework vsize, 2798981(19) pageins, 173(0) pageouts.
Networks: packets: 61207/73M in, 66123/6062K out. Disks: 877183/26G read, 171220/22G written.

PID   COMMAND      %CPU  TIME     #TH  #WQ  #POR #MREG RPRVT  RSHRD  RSIZE  VPRVT  VSIZE  PGRP PPID STATE    UID  FAULTS    COW    MSGSENT   MSGRECV   SYSBSD    SYSMACH   CSW      PAGEINS
1530  mdworker     17.1  06:56.12 4/1  2    82+  384+  39M+   16M    143M+  196M+  2586M+ 1530 1    running  501  1946213+  727+   881948+   219072+   6484219+  676187+   679401+  1914+
0-    kernel_task  14.0  07:55.29 69/4 0    2    757-  22M    0B     130M-  75M    2355M- 0    0    running  0    26279     0      33251333+ 32358872+ 0         0         10243109 0
1522  mds          12.6  14:55.33 9    8    118- 862   361M+  6052K  239M+  1334M+ 3736M+ 1522 1    sleeping 0    13373455+ 1286   12149111+ 6697348+  16039295+ 6938357+  3331251+ 1819924+
[...]

This is an example of a problem (no I don't have the same problem ☺) where my Mac was overheating like a fool. Here the 2 processes involved are mds and mdworker and are caused by Spotlight indexing an external drive of 500 Go of backups which I just plugged on. The 2 involved tasks regularly topped to 200% cpu alltogether. Spotlight is very inconspicuous about this behind the scene heavy business. It does just display a small point at the center of its looking glass within the menu bar.

When I finally clearly saw what was happening under the hood, I decided to let spotlight finish its job, and one hour later, the temperature, the processors and the fans returned to a normal life.