tl;dr: I edited this answer because it's not true; it was a quirk on my laptop which caused the low write speed of 2MB/sec after a Restart! Therefore I do not know when that refreshing of SSD cells triggers, I thought I knew but I was wrong!
Fortunately for 840EVO owners(and unfortunately for me for being wrong), further testing revealed that the write speed slowdown to 2MB/sec was NOT because of the firmware transparently refreshing the SSD cells, but instead it is a quirk on my Lenovo Ideapad Z575 laptop! (more details after next paragraph)
Therefore this answer is no longer valid and doesn't answer OP question as to when does the refreshing trigger! But I believe it's important to keep it here instead of deleting it all, so that whoever's already seen it can see the correction! Perhaps only the higher SSD temperature shows when refreshing happens: something like 44-47 degrees celsius would be an indicator(I've only once seen 56 celsius btw!), but something like 32-35 celsius would be normal no-in-the-background activity temperature. Either way I do not see a decrease in read/write performance (aside for my laptop-specific quirk).
So what happened that caused 2MB/sec random write speed for me with Lenovo Ideapad Z575?
Well, it turns out, that if I put my Samsung SSD on the optical disk drive bay (ODD) eg. via a drive caddy, then ONLY a Windows 7 Restart will trigger a mode(or something) whereby the 4KiB Q1T1 write speed reported by CrystalDiskMark is 2MB/second instead of the usual 32MB/sec. I do not know why this happens, but it can be stopped either by a system Sleep(and wake up) or by a Shutdown(ie. poweroff) and it's always triggered by a (Windows 7)Restart, but never triggered by BIOS doing a Restart for example because you changed something in BIOS and chose Save&Exit.
Furthermore, if I put the Samsung SSD on the ESATA port, not even Windows's Restart(nor Linux's restart!) will trigger this, and thus the write speed is never lowered to 2MB/s.
All this time no other SSD/HDD are connected.
Now if I add a Kingston SSD on the ESATA port and the Samsung one on ODD, then in Linux, it's the Kingston SSD that's on ESATA, and not the Samsung one on ODD that's experiencing the 2MB/sec low write speed(and only after a Linux restart eg. reboot)! This is how I know it must be a Lenovo Ideapad Z575 quirk and not a Samsung SSD issue! (I used this fio-cdm script in Linux to test the write speeds)
If I put Samsung SSD on main drive bay, the Windows Restart again has no effect on the write speed, regardless of how many restarts I do!
EDIT: I've further tested that if I put the Kingston SSD alone on the ESATA port (without any other drive on main bay or on optical disk drive bay), then after the first Linux reboot, the write speed is down to 2.7MB/sec (from the usual 34MB/sec which was present in the first boot). This further confirms that the issue is specific to the laptop!
Follows the old(wrong!) answer:
From my experience, the second boot will trigger this refreshing to happen. In other words, the first Reboot from your OS will trigger it; eg. you turn on the laptop, boot into your OS, test SSD write speed (Crystal Disk Mark's Random 4KiB Q1T1, 1 try of 50MB file) and it's normal (33MiB/sec, more or less); then you do a Restart, so the OS boots again(ie. the second time), now you test SSD speed and it's 2MiB/sec random write speed. I tested this twice. It seems consistent. If you enter BIOS and it does a reboot for you (eg. because you saved changes) then the first boot into your OS that triggers it. I guess it's all about how many ATA HDD reset commands has the SSD seen since power on: if >= 2 then starts the refresh. (TODO: test this in linux with hdparm at some point)
Another OS restart will not stop the refreshing (or if it does stop it, it starts again, so I can't tell that it stopped). The only way to stop the refreshing of cells, which degrade write performance while it is happening, is to shutdown(ie. power off) the OS/laptop. Then you start it up again and it's normal, until your next reboot which obviously triggers the refreshing.
So if you want to preserve lifespan AND write performance, just never reboot/restart, always just shutdown instead.
EDIT: My answer assumes that a Restart/Reboot does not power off then on the system(laptop or desktop) which is what actually happens on a desktop motherboard like ASUS PRIME Z370-A on which I haven't tested this SSD.
(I instead assumed, in the above post, that a Restart/Reboot keeps the system turned on, without any brief power off interruption)
I was wrong about all this strikedthrough above!
1This would all appear to be "I know better than the manufacturer how to look after my hardware." What makes you think that this change isn't an improvement in behaviour? EVOs have always suffered from bad slowdown if you never refresh them - this would appear to be taking that refresh out of the hands of users who never do it & making it automatic. – Tetsujin – 2018-09-20T18:34:56.760
@Tetsujin to be clear, after the refresh is done, the read speeds are indeed improved just as mentioned in the articles! I just don't want to pay the cost incurred during the refresh: SSD lifetime decrease, low write speeds during the refresh(which could take hours to complete, from my experience), increased frequency of 1to2minutes-OS-freeze occurrences due to 1TB of data having been written since the last such freeze (on 1TB SSD drives, at least) (this is somehow detected internally, not reported by SMART data as LBAs written) which happened only once a month(at most) with the old firmware. – None – 2018-09-21T05:46:04.117
The
– None – 2018-09-22T12:43:34.237low write speeds during the refreshthat I mentioned in the above comment was due to a quirk that happens on my laptop after Restart and completely unrelated to Samsung SSD, lookslike. So I wanted to correct that! The rest I stand by though.1
@Tetsujin given that the slowdown to 2MB/sec in write speed was for a completely different reason than the SSD firmware refreshing its cells, I can now say that I concur with you.
– None – 2018-09-23T20:20:10.183