IRQ sharing has been a thing since ACPI/APIC/PnP, but was a problem occasionally with certain devices on 33MHz PCI slots. Devices would have to compete for priority, and low-latency requirements between competing devices could potentially cause problems.
I had an IRQ sharing issue between a dial-up modem and a sound card back in the day. It was a crappy Systemax with a pentium 3. If I connected to the internet, then tried to play music, the modem would suddenly disconnect. PITA! I did what you did, switched the modem to another slot, and the problem went away.
Nowadays, it's not really an issue as PCI-E slots are so much faster, any competing interrupt requests would get resolved in a matter of nanoseconds. It's likely that the PCI-E spec itself, in addition to being faster, also has some other mitigations baked in.
IIRC a way to test for potential problems is using tools like DPC Latency checker/LatencyMon, but I think those more relate to problems stemming from drivers.
It would be interesting to test if there's any perceived change in performance/latency when high-bandwidth devices have to share an IRQ vs. when they don't, in modern systems.