I have heard that a situation where a 24V Gbit Ethernet port was not completely fried after being exposed to 48V current, but downgraded to fast Ethernet - 100Mbps.
Question: I that possible and how?
POE over gigabit and POE over fast ethernet have the potential to be different.
Ethernet over twisted copper of any kind utilizes "differential signaling" - meaning that the waveform that provides data for each twisted pair is actually juxtaposed against another reference "carrier wave" rather than relying on the difference between the data signal and ground. This allows both of those waveforms to be "pushed up" by X amount of DC voltage without affecting the data signal - and consequently allows every copper conductor in a gigabit line to be used for power as well as data. This is referred to as "Alternative A POE", and is required in gigabit implementations (gigabit requires the use of all eight conducts for data, so we can't dedicate half of them to only power).
Fast Ethernet doesn't use as many conductors (half, for a total of four), so POE implementations can be run on the "other half" of the cable without employing a DC bias in the data lines. This is referred to as "Alternative B".
As far as why that situation you heard of was possible, there are many devices that can autonegotiate POE as well as accept 24V or 48V. Sometimes, that autonegotiation involves using a different 48 volt circuit that is only accessible using fast Ethernet standards, but that wouldn't be terribly common - however it's a thing I've run into in the past with some very old (circa ~2000) devices. In general, this autonegotiation should just work on modern devices regardless of speed. However, you should ALWAYS refer to the device documentation to make sure that all of your equipment will correctly auto-negotiate.
IEEE PoE ("DTE power via MDI") only uses 48/54 V voltage and is carefully designed to not damage any legacy equipment. However, proprietary, non-standard PoE may not interoperate well with IEEE PoE and may also cause damage to incompatible equipment. You should always prominently mark these ports and keep them well apart from the rest of the network.
With proprietary variants, it's theoretically possible to fry part of the magnetic coupling - the pulse transformers - reducing the working pairs in a link. Some gigabit ports fall back to 100BASE-TX ("Ethernet@Wirespeed") when they detect non-working pairs, so yes, it's possible that a partially fried 1000BASE-T port continues to work at 100 Mbit/s. The fallback may even happen on the non-fried side, so the port may only work with certain combinations. Typically, falling back to 100BASE-TX takes a couple of seconds and the link will take more than the usual time to come up.
Using a simple multimeter on an unpowered port, you should be able to measure 2-5 Ω across each of the pairs (1-2, 3-6, 4-5, 7-8 for 1000BASE-T, with 10BASE-T and 100BASE-TX only using 1-2 and 3-6). An open circuit or 0 Ω means the transformer is fried. (The AC impedance is 100 Ω but a multimeter uses DC.)