Rated current will have a long holding time max (hrs) and wide tolerance for min, so they are rated like \$0.02 A^2t @ 0.2A\$ at 10xI
E.g. at 10 x 0.2A rating , =2A, t=0.02/A^2 = 0.5s fuse time.
Melting I2t 10.0 In typ. [A2s] for 0.2A @ 10x rating.
Thus accuracy of time improves with overcurrent load and not reliable is loaded just at fuse rating. So you choose a rating slightly over demand current averaged over a period of time like 150%.
What this means is fuses get warm with about 1/4W loss just below rated current, yet not enough to fuse it. Depending on the ambient and filament temp, exceeding this current above rating causes the temperature and due PTC also resistance, which in turn accelerates power loss in a \$I^2t\$ quadratic function thus reducing the fuse time. This is called thermal runaway but is slow at rated current in hours or less then accelerates with this quadratic characteristic.
Because the mass is small, it can only protect devices from damage due to thermal rise damage in a similar fusing quadratic characteristic and not say for semiconductors with wirebonds much smaller than the fuse unless of course rated higher than the fuse. In those cases, component protection demands other means for abrupt current cutoff like an open switch or a “crowbar” shorting the fuse with some current limit.
For example gastubes are often used as a over-voltage protection (OVP) crowbar and when triggered, drop in voltage to something like 60V across the grid source and need a fuse between gas tube and circuit to prevent melting a hole the board. ( seen that before) But that can protect the circuit from the following grid current due to a voltage breakdown on the power supply, triggered by lightning.