good question, what do others think?
personal view, i can't see any issue.... unless we start getting a huge number of entries, when it might be a pain during the voting process. but for now Id say the more the merrier 
time.... but apart from that, I guess it depends cpu if doing large number of voices, sram if not...
(i do try to keep it in check as i go along)
sram, make sure if you are doing lots of subpatches to zero out presets/modulation sources unless you are using them, they can consume alot of memory... which you may not need
SRAM meter, I think thats a bit tricky, I think, we only know when we compile and link the patch... so we would have to be background compiling to do this, which would have cpu load... and the have to interpret the errors. could be possible in the future (optional in case you are using a slow computer) .. the issue though is this can only say when it fails.
hmm, unless we perhaps read the elf file? ... hmm perhaps that could work, if we report each time to went live... then at least as you iterate through development you would see it grow...
alternatively I guess we could guesstimate, by looking at the sizes of objects... and knowing the constant overheads e.g. parameter maps... but probably hard work, and not very precice.
