ExtremeTech came across an interesting report that investigates the reliability of solid state disks. Only a small subset of SSDs were tested, and specifically disks with limited storage capability. During the test, the disks were put through repeated power cycling, massively parallel write operations and direct disk writing. The difference in reliability seems to be huge, the OCZ disk for instance failed the massively parallel write test, even with a steady power state, but passed the test after downgrading the firmware to an older version that wasn't optimized for higher performance. The test also illustrates that many disks have issues with power cycling during the read-synchronize-write cycle: after 1600 power cycles, the Crucial M4 was recording up to 40,000 CRC errors. Only the Intel 320 and S3500 didn't suffer from data corruption, despite being power-cycled over 6,500 times.
Just power cycling the drives while no read/write operations were occurring was no problem, but power cycling them during the read-synchronize-write cycle was incredibly problematic. After 1600 power cycles, the M4 was recording up to 40,000 CRC errors. The Toshiba THNSNH060GCS upgrade kit was able to maintain file integrity if file writes were handled at less than 20MB/sec total, even when writing 64 threads of data. Exceed that rate, however, and the Toshiba drive starts losing data quickly.