For a long time I've had some issues with packet loss and Rx FCS errors on the other end of the fiber.
The weird thing is, the issue depends on the number of SFP modules inserted (even if not linked) close to each other, and doesn't seem to be thermal (also happened at lower temperatures in winter), more like power or signal integrity issues.

CRS328 sfp1 - Tx1550/Rx1310 SC WDM SFP - 300m of SM fiber - Rx1310/Rx1550 SC WDM SFP - NetPower Lite 7R SFP+2
or
CRS328 sfp1 - 3m copper DAC - CRS105-5S-FB sfp2 (sfp1 DHCP client for management only, sfp2-sfp5 as HW offloaded bridge with minimal config)

In both cases I can see Rx FCS errors on the remote end (NetPower Lite 7R or CRS105-5S-FB) and a few percent packet loss, but only if there are 5 other optical SFPs (just inserted, even with no fiber connected) in nearby SFP cages within the same 8-port group, not with just one other SFP inserted.

Could this be a bad port, or common PHY chip for 8 ports? No, moving from sfp1 to sfp9 fixes the issue... until the other 5 SFPs (without connected fiber) are moved to sfp10-14.

This only affects the 1G SFP ports, the same device has all 4 SFP+ used and loaded with heavy traffic with no issues so far. And it's the newer (presumably better) "r2" HW revision. No difference with with RouterOS 7.16.2 or 7.19.1. Has anyone else seen something like this?

BTW, it's worse with Netpower Lite 7R as after each such RX FCS error it not only loses the bad packets, but seems to reset the port, gets IP via DHCP again like after link down/up events, and probably loses a few more packets in the process. It took quite a lot of time to debug, and the CRS328 was the least suspected device (after all, the hardware not cheap so should be of high quality, and 10G SFP+ ports work well so why lowly 1G SFP wouldn't?).