More generally, no component is pure; once you understand that a resistor is an RLC network like any other component, it becomes not a question of "best resistor for X", but "which resistor has the most bandwidth / flattest impedance / least EPC/ESL for X application".
And then you understand ways to game the system: you can increase the bandwidth/flatness of a desired resistance by using series/parallel combinations of parts with wider bandwidth. The bandwidth is invariant (it can't be greater than the part's EPC/ESL and R determine) under series/parallel combination, so if you start with wider band parts, you can potentially improve things. Example: maybe that 6.2M resistor could be made from six 1M resistors and one 200k, which will each have the same EPC (effective parallel capacitance) as a single 6.2M, but six in series is 1/6 the capacitance. Mind I said BW can't be greater than the part's own property; it can always be lower due to circuit strays. This isn't a slam-dunk solution, it's just an option in your toolkit. (Scope probes constructed this way, usually need to deal with the distributed stray capacitance along the resistance element / divider chain -- sometimes this is done by swamping the stray capacitance with much more external capacitance, sometimes this is done by nulling cancelling out poles and zeroes in the frequency response. Basically, this is why some scope probes have multiple trimmers on the probe or plug body.)
So, if you ever wanted to know much, much more than you thought you wanted; this is one of many ways to get there.
As for OP's exact situation, it's very likely that a direct replacement, respecting voltage and power ratings, will suffice. A large value like that, is suggestive of a bleeder application (often, high-value resistors are used line-to-line to discharge capacitors of nuisance voltage; the value is noncritical), or a bias resistor (the value should be accurate), or something for high voltages. Do check around for connected high voltage components -- if present, it should probably be replaced with a high voltage part, or at least a series chain of regular rated parts. (Often, carbon-comp resistors were abused in this way, and their values typically drift high when operated at high voltages!) It's very unlikely that the resistor's bandwidth matters, but if that is the case, some compensation is probably necessary (you'll have to inspect the circuit much more closely to figure how to do that).
Tim