Apple SSD Disconnects During Recovery: Image First or Stop DIY?
Stop treating this as a partition-repair problem. An SSD that repeatedly ejects during a byte-to-byte copy, disappears from Disk Utility, and returns only after a reboot should be treated as an unstable source. If the remaining data matters enough to pay for, stop powering it and move to a professional data-recovery lab. If you knowingly accept a one-shot DIY risk, the next goal is a sector-level image—not more First Aid, Finder copying, or file-system repair.
That is the state reached in this Early 2015 MacBook Pro recovery case. The owner had a flashing-folder startup failure, repeated Disk Drill/SuperDuper copy failures, First Aid errors, and an internal Apple SSD that ejected during reads. They later managed to copy the most important bookmarks and text files through Terminal, but the Macintosh HD volume then stopped appearing again.
What The Repeated Ejection Changes
Apple says a flashing question mark means the startup disk is unavailable or no working macOS can be found. Apple also recommends service when the startup disk does not appear in Disk Utility or cannot be repaired. In this case, the more important signal is not the question-mark icon itself but the repeated loss of the SSD during sustained reads. Apple's startup-disk guidance supports treating non-detection as a service-level condition rather than continuing normal repair attempts.
Do not let a reported “SMART OK” result override repeated disconnects and I/O errors. The source has already shown that it cannot stay online reliably enough for ordinary copying.
If You Accept One DIY Attempt, Image The Source First
GNU ddrescue is designed to copy readable areas from a failing device first and postpone difficult regions. Its manual explicitly warns against repairing a filesystem on a drive with I/O errors and recommends making a copy before working on the filesystem. GNU ddrescue's official manual is the right reference if you choose this path.
- Step 1
If the SSD reappears as a block device, do not mount it read-write or run First Aid again. Prepare a healthy destination large enough for a full image or clone.
- Step 2
Use an image-first tool intended for unstable media and keep a map/log so the session can resume after a disconnect. Stop if the drive becomes harder to detect or you are no longer willing to risk the remaining data.
- Step 3
Only after a stable clone or image exists should you try file-system or partition tools. TestDisk is designed for logical partition/file-system work; its documentation describes rewriting partition structures, which is not what you want to do first on the unstable original.
Decision Point:
Because the most important files were already copied, decide how much the remaining data is worth before the SSD gets another power cycle. If losing the rest would be unacceptable, stop DIY now. If you accept that the next read may be the last, image first and perform all later recovery work on the copy.